Exhibit 4.75

English Translation

Contract Registration No.:

2 0 2 2 1 1 0 0 3 2 0 0 3 4 0 2

Technical Service Contract

Project name	Business Technology Application Solutions Based on Big Data Analysis
Service Recipient (Party A)	Beijing Sohu Donglin Advertising Co., Ltd.
Service Provider (Party B)	Beijing Sohu New Media Information Technology Co., Ltd.

Place of signing: Haidian District, Beijing City

Date of signing: January 1, 2022

Validity period: January 1, 2022 to December 31, 2024

Beijing Technology Market Management Office

- 1 -

Whereas Party A (Service Recipient) desires to entrust Party B (Service Provider) with the provision of technical services for the project entitled Business Technology Application Solutions Based on Big Data Analysis;

Whereas Party B is willing to accept the entrustment of and provide technical services to Party A;

In accordance with the provisions of the Civil Code of the People’s Republic of China on technical contracts and other relevant laws and regulations, through amicable negotiation, the parties agree to enter into this contract (the “Contract”) based on the following terms and conditions, and to abide by them.

1 Content, Methods and Requirements of the Service

1.1 Content of Technical Services as the Subject Matter of the Contract

(1) Definition of Relevant Terms

① Online Advertising Agency shall mean Party A of the Contract whose primary business is to contract with and expand advertisers in need of Internet/mobile Internet marketing services.

② Technical Service Provider shall mean Party B of the Contract which provides technical support for digital upgrade of enterprise marketing, such as online advertising putting, optimization, monitoring and evaluation.

③ Advertisers shall mean all the existing and potential business customers of Party A hereof.

④ Business Technologies shall mean the Internet/mobile Internet advertising marketing technologies owned by Party B with independent intellectual property rights.

(2) Project Overview

Amidst the wave of digitalization, online advertising industry is undergoing rapid transformation in both production and dissemination modes. Consequently, digital upgrade of marketing has become an increasingly obvious trend as digital means are used to build and connect various marketing processes, and interaction of full-chain touch points and conversion of key touch points become major demands of the advertisers. Meanwhile, security of data and personal information has received attention at the country level. This shows that with accelerating social digital transformation, data has become an important asset promoting the application of marketing technology.

- 2 -

Through an analysis of advertisers’ needs in the post-epidemic era, it is clear that driven by the epidemic and wave of digitalization, their cognitive and marketing strategies have changed to focus more on precision marketing and cost control, with higher investment in digital technology. In the meantime, with continually optimized enterprise structure in response to the epidemic, recovery of the external environmental economy, extensive application of digital technology, and the maturing of more diversified marketing forms, the advertising sector has entered a stage of refined and efficient development.

The above analysis on the status quo of China’s online advertising business shows that new marketing technologies, as compared to traditional advertisement putting and activity planning, feature clear indicators and billing methods. In the context of rising marketing costs and increasing customer acquisition costs (CAC), advertisers hope all the more to quantify marketing inputs and outputs to the maximum by applying new marketing technologies so that a breakthrough can be made in marketing efficiency. Therefore, the goal of the technical services hereunder can be summed up as follows: Through innovative application of the digital technology, Party B shall propose business technology application solutions based on big data analysis so that advertisements can be injected more accurately and efficiently, and marketing value can be maximized by addressing the actual problems of consumers. Eventually, Party B shall provide proper online marketing technologies for Party A to drive digital upgrades of its existing and potential advertiser clients.

Business technology application as mentioned herein focuses on resolving the problems that advertisers want to resolve the most, including intelligent and transparent advertisement putting, automatic traffic input and calculation and building of a private domain operation matrix, and improved work efficiency through connectivity with digital business of other departments. In the course of providing technical services to Party A, Party B will mainly use the following core proprietary business technologies.

① A method and system for calculating attention

This technology relates to technical field of data processing.

With the development of the Internet, masses of reading data are generated as users browse news, journals and other articles. By analyzing such reading data, we will be able to understand specific content of a particular industry that a user is interested in. According to the existing method of analyzing reading data, keywords are extracted from the articles a user read; based on the frequency of occurrence of each keyword, the user’s interest in specific content of a given industry can be determined. However, keywords of a certain sector may appear in articles covering diverse fields and have different influence when used in different fields. Therefore, the existing way of analyzing reading data is not accurate in determining what specific content of a given industry is of interest to the user.

- 3 -

Based on the number of views of each article containing the keywords to be analyzed and the author’s influence, this technology application calculates user attention to such keywords by considering their weight in each article. Utilizing the frequency of these keywords, their weight in different articles and the author’s influence, user’s attention to such keywords can be calculated more accurately.

② A method and device for predicting crowd flow

This technology relates to the technical field of data processing.

When putting advertisements, advertisers set directional conditions to obtain accurate target crowd flow. The finer the directional conditions, the more accurate the flow obtained and the smaller the crowd coverage, which means less range of exposure. Therefore, advertisers need to strike the balance between obtaining accurate target crowd flow and ensuring certain range of exposure. To do this, they set combinations of directional conditions and predict the crowd flow they can bring.

This technology application can provide a flow prediction method and device to quickly predict the flow of certain label data.

③ A method and system for determining information similarity

This technology relates to the technical field of data processing.

A relatively common information recommendation algorithm in the information recommendation area is collaborative filtering algorithm. When recommending information using the collaborative filtering algorithm, it is necessary to calculate the similarities between information. The current method of calculating information similarity using the collaborative filtering algorithm determines information similarity based on the number of times they appear together. In practical application, however, lots of sparse user behavior and few information co-occurrences can result in less information coverage by the collaborative filtering algorithm, and consequently lower recommendation accuracy.

This technology application can provide a method and system for determining information similarity. By increasing the amount of information covered by the method of calculating information similarity, it improves the accuracy of information recommendation.

④ Product recommendation method and device

This technology relates to the technical field of data mining.

- 4 -

With the spread of the Internet, more and more users start to use online shopping platforms (“online platforms”). They can view product (including both virtual and physical products) information on the webpages of the online platforms, and select and purchase some of the products. To aid the selection process, online platforms generally provide the product recommendation function, which predicts user preferences using the recommendation algorithm and recommends products to users as they visit the online platforms.

Collaborative filtering algorithm is a type of exiting recommendation algorithm. Its accuracy depends on the length of the real behavior sequence (referring to the number of products recorded in the actual behavior sequence) used for calculation. For users with multiple short real behavior sequences, the collaborative filtering algorithm is less accurate in determining the similarities between users, and thus cannot recommend products accurately.

This technology application can increase the length and co-occurrence frequency of the behavior sequence for candidate users, thereby improving the accuracy of product recommendation.

⑤ Video recommendation method and device, storage medium and electronic device

This technology relates to the technical field of data processing.

As Internet technology develops rapidly in the past few years, video platforms are attracting an increasing number of users. All kinds of videos which contain masses of fun and rich contents have emerged as an important recreation in people’s daily life. However, the deluge of videos on the Internet has made it hard for users to quickly choose the ones they are most interested in. To enable the users obtain the videos they like most from huge amounts of videos, the existing technology generally identifies videos similar to those watched by a user by their titles, key frames or videos, or in other ways. However, in the absence of a viewing record, there is no way to identify the videos that are of interest to the user.

This technology application can recommend videos to users even if they do not have a previous viewing record.

⑥ A video recommendation method and its target service provider, service caller and system

This technology relates to the technical field of video recommendation.

- 5 -

The development of the Internet has put an increasing number of video software on the app market. During the application of video software, user characteristics and video features are used to recommend the corresponding videos to users. Existing recommendation is based on the click rate of videos obtained by the machine learning model using user characteristics and video features. Nevertheless, as videos increase in both quantity and length of feature, it takes relatively long time to process the videos before they can be recommended to users. Besides, videos cannot be processed in large quantity. Low video processing capacity has further led to a delay in video recommendation.

This technology application sets up multiple service providers and adopts the load balance strategy to select the corresponding target service provider for video recommendation processing. This improves the video processing capacity and thus reduces delay in video recommendation.

⑦ A method and device for information recommendation

This technology relates to the technical field of information recommendation.

In practical application, the deep factorization machine (DeepFM) model is adopted to predict the click rate of information. When such model is used to estimate the click probability, the dimension of the model must be increased in order to make the prediction more accurate. This way, the DeepFM model would consume large computing resources; it also takes a long time to estimate the click rate of information. In brief, this method of predicting information click rate can hold large computing resources while greatly delays the prediction process.

This technology application predetermines and caches the cache vector of splicing characteristics for each piece of information in the information base. When using the DeepFM model to predict click rate, the corresponding cache vector is found directly from the cache for click rate prediction. While ensuring the prediction accuracy of the DeepFM model, this reduces the computing resources it holds and hence reduces the delay in click rate prediction.

⑧ The method of recommending multimedia information, related device and computer storage medium

This technology relates to the technical field of item recommendation.

- 6 -

In order to promote and sell articles, videos, commodities and other items, and to better meet the needs of users, today we would always recommend items using the recommendation algorithm, and primarily, the collaborative filtering algorithm. Because this way of recommendation relies on historical interaction data of users, it applies only to items that appear together. When two items do not co-occur, it would be impossible to identify their similarities or even recommend items based on such similarities.

Given the deficiency of the existing technology mentioned above, this technology application can address the challenge of calculating similarities for items where there are no user co-occurrence interactions. In other words, even if two items lack the interaction of user co-occurrence, their similarities can still be calculated accurately using this technology and eventually, items can be recommended to users based on such similarities.

1.2 Mode of Technical Services

It is agreed by Party A and Party B that the technical services (subject matter of the Contract) hereunder shall be provided in the manner below: Relying on its experience in the development and management of Internet/mobile Internet software platforms as well as in information promotion operation & maintenance, Party B (Technical Service Provider) shall utilize its proven and proprietary information technology for Internet/mobile Internet promotion services to provide Party A with online marketing technology solutions (i.e. business technology applications based on big data analysis), thus assisting the Service Recipient (Online Advertising Agency) in meeting the needs of existing and potential advertiser clients for digital upgrade of online marketing.

1.3 Requirements of Target Technology

1.3.1 Technical Requirements

In order to achieve the goal of the technical services hereunder, this Project is committed to creating innovative network products and marketing tools and bringing about diverse advertising and marketing ideas through constant innovation. The ultimate purpose is to enhance the audience-brand interaction, and boost the brand value of advertisers.

For the detailed design of technical solutions and applications relating to the “business technology applications based on big data analysis” mentioned in the Contract, please refer to the Annex Technical Service Scheme Specification.

1.3.2 Performance Requirements

(1) Technology: New marketing technologies empower as datamation enters a difficult phase

- 7 -

Against the background of ever-improving data technology, strengthening customer data management and achieving precise marketing operation have become the biggest concern of future advertisers. Meanwhile, private domain operation has also been favored by many advertisers as its characteristic value is constantly being explored.

Here are the problems that advertisers desire to resolve the most: intelligent and transparent advertisement putting; automatic traffic input and calculation, and building of private domain operation matrix; and improved work efficiency through connectivity with digital business of other departments. While new marketing technologies are still at the early stage of application, their further penetration is expected to fully empower online marketing.

(2) Strategy: Creative content and experience becomes the next trump card to play

The development of the Internet has prompted the continuous emergence of new economic forms and at the same time, brought about tremendous changes in the advertising and marketing pattern. As the preferences of advertisers shift from traditional 4A creation and production to the mindset of traffic and refined operation, they start to place a higher value on the accuracy and quantitative effect of advertisement putting.

In the years to come, as advertising technologies continue to upgrade and advertisers’ demand for results-oriented marketing are satisfied, they may well turn to an advertising form where technology and content are highly integrated. The content marketing strategy which gains momentum recently for its richer and more in-depth information, and stronger emotional resonance generated from closer combination with content, is drawing more and more attention from advertisers. In particular, social platforms and short-video platforms that are most tightly bound with content, are appreciated by most advertisers.

(3) Scene: Media integrate and interconnect to achieve full scene coverage

With the further penetration of digitalization in various industries, increasingly diversified online and offline marketing channels, ever-upgrading consumer demands, and constant expansion of retail scenes, it would be increasingly difficult for a single medium to meet advertisers’ marketing needs and consumers’ information needs at the future environment of information overload. Therefore, strengthening the Internet nature of all types of media, breaking down the barriers between them, and achieving coverage of all consumption scenes and deep interaction of user systems have become the mainstream of future marketing development.

- 8 -

In the field of digital marketing, future advertisers look forward to establishing a rich and perfect full-channel system. By applying intelligent decision-making, they aim to fully connect with consumers through online and offline touch points for targeted marketing.

(4) Concept: Upgrading and innovation from the point to the volume mode

Driven by the general trend of digitalization, digital transformation in enterprises are accelerating and deepening. Based on current developments, the transformation process can be divided into four phases: point, line, area and volume. Digitalization starts with the application of a single digital technology, and evolves further to the optimization of overall business process and building of a trans-department digital platform until eventually, the barriers of traditional business ideas are broke, and innovation and reform of enterprise’ marketing models are realized.

In the process of digital upgrade, marketing as a core business unit will further strengthen its connection with other units, such as brand, product and finance. Traditional division of single business segment is replaced by an operation concept that encourages interconnection and coordination among different units. As a result, the marketing department is upgraded to be a digital-driven marketing team, channel team, or customer operation team. Through data collection, calculation, application and reflux in each step, marketing technologies are applied throughout the marketing process. With the deepening of the digital concept, problems that are inherent in traditional organization structure including isolation between business segments, unclear rights and responsibilities, and high communication costs have been improved. Business processes have been institutionalized with quantifiable and attributable results, and well-defined KPI assessment criteria.

According to the above analysis on the development trend of new online marketing technologies, as the epidemic speeds up digital transformation across the society, new marketing technologies have entered a difficult period of development. By implementing the technical services hereunder, the company strives to achieve the following goal: By applying the big data technology, we aim to promote the digital upgrade of marketing and make digitalization part of the whole marketing process, so as to add new momentum to the online advertising market and explore new business models as the digital concept continues to deepen and innovate.

2 Working Conditions and Matters of Cooperation

For the Service Provider (Party B) to carry out its work smoothly, the Service Recipient (Party A) shall provide the Service Provider (Party B) with necessary working conditions and cooperate with Party B technically.

- 9 -

Party A shall provide the following working conditions and cooperation:

2.1 Party A shall specify the personnel to involve in the project, and make time and work arrangements in such a manner as to meet Party B’s requirements for performing the technical services.

2.2 Party A shall clearly express its demands in writing, and provide related technical background materials, technology, data, original design documents and necessary samples.

2.3 Party A shall provide additional explanations, data and information as required by Party B.

2.4 Party A shall promptly modify and improve any technical materials and data provided to Party B that contain obvious errors and defects.

2.5 Party A shall provide the work places for Party B’s personnel to perform the services at Party A.

2.6 Party A shall provide necessary facilities and equipment to ensure the provision of services by Party B.

2.7 On the premise that Party B’s work meets relevant requirements, Party A shall make sure to carry out acceptance inspection in accordance with the corresponding procedure. This includes providing Party B with personnel, content, procedure and the environment, and preparing other necessary conditions.

3 Term, Place and Method of Performance

3.1 Term and Place of Performance

3.1.1 This Contract shall be performed in Haidian District, Beijing City from January 1, 2022 to December 31, 2024.

3.1.2 During the term hereof, both parties shall continue or require their successors in rights and obligations to abide by and perform their respective obligations hereunder, regardless of any change in the name, organizational form, enterprise nature, business scope, registered capital, and investor of each party.

3.2 Method of Performance

- 10 -

The parties agree as follows regarding the performance of the Contract: Relying on its experience in the development and management of Internet/mobile Internet software platforms as well as in information promotion operation & maintenance, Party B (Technical Service Provider) shall utilize its proven and proprietary information technology for Internet/mobile Internet promotion services to provide Party A with online marketing technology solutions (i.e. business technology applications based on big data analysis), thus assisting the Service Recipient (Online Advertising Agency) in meeting the needs of existing and potential advertiser clients for digital upgrade of online marketing.

4 Criteria and Method of Acceptance Inspection

In the fair, scientific and practical principle, the parties hereto agree that the technical services hereunder shall follow the iterative development standard. Services shall be accepted once product functions are completed by phase and can be released. When Party A issues a service acceptance certificate, it is deemed that the services have passed the inspection and are accepted by the Service Recipient.

The warranty period for the services hereunder shall be 5 working days after the issuance of a service acceptance certificate by Party A. If any service is found to be defective during the warranty period, Party B shall be responsible for reworking or taking remedial measures unless the defect is due to improper use or care by Party A.

5 Remuneration and Its Payment

5.1 By referring to the service fee standard of the industry and its payment method, both parties agree that remuneration for the services hereunder shall be paid through revenue sharing, namely, Party A shall pay 90% of the operating revenue from the technology platforms hereunder to Party B as the technical service fee.

5.2 Party B shall bear relevant expenses necessary for completing the services hereunder during the term of the Contract, which include but are not limited to costs of survey, investigation, analysis, demonstration, tests, determination, and other activities necessary for the development and promotion of product technologies required for providing the service, equipment depreciation costs, broadband rental charges, overhead expenses, market development expenses, etc.

-11-

5.3 Given the use of Party B’s proprietary technology of independent intellectual property rights during the provision of technical services hereunder, through friendly negotiation, the parties agree that Party A shall pay royalties for the following related technologies for which Party B has obtained a patent for invention when making the first payment of remuneration to Party B:

① A method and system for calculating attention: RMB 2 million

② Product recommendation method and device: RMB 2 million

③ Video recommendation method and device, storage medium and electronic device: RMB 3 million

④ A method and device for predicting crowd flow: RMB 2 million

⑤ A video recommendation method and its target service provider, service caller and system: RMB 3 million

⑥ A method and device for information recommendation: RMB 2 million

⑦ A method and system for determining information similarities: RMB 2 million

⑧ Multimedia information recommendation method, related device and computer storage medium: RMB 3 million

The aforesaid royalties (inclusive of tax) add up to RMB 19 million.

5.4 Payment method: It is agreed that Party A shall pay the service remuneration to Party B by check or through bank transfer.

5.5 Time of payment: The remuneration payable by Party A to Party B shall be settled monthly. After the calculation of such remuneration is confirmed by Party B, Party B shall issue a special VAT invoice to Party A, and payment shall be made by Party A to Party B during the first 10 working days of the following month.

5.6 Minimum return clause: Party B undertakes to ensure a ROC (return on capital) of no less than 20% for Party A during the valid term hereof.

5.6.1 Definition of ROC

Party A’s ROC = Party A’s Annual Net Profit / Paid-up Capital×100%

5.6.2 Method of implementation: Before paying any remuneration to Party B, Party A shall conduct financial accounting in accordance with the revenue-sharing ratio agreed in the preceding paragraph. Based on the calculation result,

(1) If the ROC of Party A is equal to or greater than 20%, the revenue shall be shared according to the ratio agreed in the preceding paragraph;

-12-

(2) If the ROC of Party A is less than 20%, Party B shall make up the difference to Party A based on the minimum return clause, namely, Party A shall reduce the amount of technical service fees payable to Party B in the current period so that Party A’s ROC reaches up to 20%.

5.6.3 When Party A’s ROC is less than 20% and the minimum return clause applies (i.e. Party B is required to make up the difference to Party A based on the minimum ROC guaranteed), Party B shall have the right to require an audit of Party A’s finance for that period, and Party A shall cooperate with such request.

6 Agreement on the Liabilities for Breach and Liquidated Damages

In case of a breach of the Contract, the breaching party shall be liable for breach in accordance with relevant provisions of the Civil Code of the People’s Republic of China.

6.1 Definition and Liability of Breach by Party A

6.1.1 If Party A fails to provide Party B with relevant technical materials, data, samples and working conditions at the specified time, place and in the specified manner, it shall be deemed as Party A’s failure to give effective cooperation as agreed herein. Where the progress and quality of work are affected, Party A shall indemnify Party B for all the expenses paid for performance of the Contract and provide compensation equivalent to double of such expenses. Further, if Party A delays providing the agreed material and technical conditions for up to 20 days, Party B shall have the right to terminate the Contract, in which case Party A shall pay liquidated damages or compensate Party B for the losses caused thereby.

6.1.2 If Party B finds that the technical materials, data, samples, materials or working conditions provided by Party A are inconsistent with the provisions hereof, it shall notify Party A without delay; Party A shall supplement, modify or replace them within the agreed time limit. Party A shall bear the corresponding liabilities if it fails to reply within the prescribed period after receiving the notice.

6.1.3 If Party A breaches the Contract for no reason, or refuses to accept or delays accepting Party B’s work results, which affect the progress and quality of work, Party A shall take the corresponding liabilities and pay Party B the remuneration payable to it.

6.1.4 Party A shall pay liquidated damages and storage fees if it delays accepting any work results. Where the delay exceeds 60 days, Party B shall have the right to dispose of the work results, and deduct the remuneration, liquidated damages and storage fees from the proceeds before returning the rest to Party A or, if the proceeds are insufficient to cover the remuneration, liquidated damages and storage fees, require Party A to make up the deficiency.

- 13 -

6.1.5 During the performance of this Contract, if Party B finds that materials, samples or equipment would be damaged if work is continued, it shall stop the work and promptly notify Party A or put forward suggestions. Party A shall give a reply within the agreed time limit or bear the corresponding liabilities if it fails to do so.

6.1.6 If Party B’s work results and services are defective, but Party A agrees to use them, Party B shall reduce the remuneration and take corresponding remedial measures. If the work results and services are seriously defective that the technical problems agreed herein are not solved, Party B shall waive the remuneration and pay liquidated damages or compensate for the losses.

6.1.7 If Party A fails to pay remuneration to Party B according to the time and amount specified herein, it shall pay liquidated damages at 0.1% of the overdue fine for each day of delay.

6.2 Definition and Liability of Breach by Party B

6.2.1 If Party B fails to complete the services as agreed herein due to its own reasons, Party A shall have the right to require party B to make supplements or corrections, and Party B shall be liable to Party A for the losses caused thereby. Where the supplement or correction is delayed for up to 15 days, or the services still fail to meet the standard despite the supplement or correction, Party A shall have the right to terminate the Contract. In that case, Party B shall return the technical materials, samples and the remuneration received, and shall also be liable to Party A for all losses caused by the termination hereof.

6.2.2 In case Party B delays delivering the work results, it shall assume all the resulting liabilities, namely, Party B shall bear all the related fees already paid for the performance of this Contract and promise to pay Party A liquidated damages at an amount to be separately agreed by the parties.

6.2.3 Party B shall indemnify Party A for the losses caused if any samples or technical data submitted by Party A is lost, missing, deteriorated, contaminated or damaged due to improper care by Party B. The specific amount of indemnification shall be separately agreed by the parties when the loss actually occurs.

6.2.4 During the performance of this Contract, if Party B finds that materials, samples or equipment would be damaged if work is continued, it shall stop the work and take appropriate measures; if Party B neither stops the work or takes appropriate measures where necessary, nor promptly informs Party A, it shall assume the corresponding liabilities.

6.2.5 Party B shall be liable for the losses of Party A that arise from its breach of the confidentiality obligation agreed herein.

- 14 -

7 Dispute Settlement Method

Any dispute arising from the performance of the Contract shall be settled through arbitration or judicial procedure if the two parties fail to reach reconciliation or accept mediation.

7.1 Governing Law

This Contract shall be interpreted in accordance with and governed by the laws of the Mainland of the People’s Republic of China.

7.2 Dispute Resolution

7.2.1 Any dispute arising out of or in connection with the performance or interpretation of this Contract shall be negotiated amicably by both parties. If no settlement can be reached through negotiation, the dispute shall be submitted to and finally decided by Beijing Arbitration Commission in accordance with its arbitration rules and procedures in effect at the time.

7.2.2 The arbitration shall be conducted in the Chinese language.

7.2.3 The arbitration award shall be final and binding upon both parties hereto.

7.2.4 The arbitration fee shall be borne by the losing party unless otherwise stated in the arbitration award.

8 Confidentiality of Technical Information and Data

During the valid term hereof and the period of confidentiality agreed herein, both parties shall abide by the following duty of confidentiality, and assume the corresponding liabilities for breach of such duty of confidentiality.

8.1 Technical information and data referred to herein include:

8.1.1 Any commercial, marketing, technical, operational data or information of other nature (in whatever form and on whatever carrier) provided by either party to the other during the validity of the Contract for the purpose of completing the project hereunder, whether or not indicated as confidential orally, graphically or in writing at the time of disclosure.

8.1.2 This Contract and all the annexes and supplementary agreements hereto signed by the two parties; all the software, software catalogs, documents, information, data, drawings, benchmark tests, technical specifications, trade secrets, and other information exclusively owned by Party A or Party B and provided to the other party indicating clearly as “confidential information”, including all items defined as “confidential information” in other contracts signed by Party A and Party B both before and after this Contract.

- 15 -

8.1.3 The aforesaid confidential information can be in the form of data, text or tangible media containing such content (including materials, CDs, software and books), or transmitted verbally or through other audio-visual means.

8.2 The rights and obligations of the two parties include:

8.2.1 Each party acknowledges and warrants that it shall keep the trade secrets of the other party strictly confidential, and shall not disclose the same without obtaining the formal written consent of the other party beforehand.

8.2.2 Both parties warrant that the confidential information shall be used only for purposes related to the cooperation.

8.2.3 Each party undertakes to properly keep the confidential information provided by the other party, and to maintain confidentiality of such information as agreed herein. Each party shall also take care of the confidential information using at least the same protective measures and degree of prudence applicable to its own confidential information.

8.2.4 Either party who provides confidential information in writing shall indicate the same as “confidential”; in the case of oral or visual disclosure, the disclosing party shall inform the recipient of the confidential nature of such information prior to disclosure, and shall confirm in writing that the disclosed information is confidential within five (5) days after disclosure.

8.2.5 Confidential information shall be kept in a safe place, and the right to use confidential information related to the project hereunder shall be strictly controlled. Each of Party A and Party B warrants that confidential information shall be made known only to the persons in charge of the project and the employees engaged in services related to the project on each side. Before the aforesaid personnel are informed of the confidential information, Party A and Party B shall remind them of the confidential nature of such information and their duty of obligation toward such confidential information. Further, Party A and Party B shall make sure that they agree in writing to be bound by the terms of this Contract and undertake confidentiality responsibility to a degree no less than that specified in the Contract.

8.2.6 At the request of the disclosing party (the “Disclosing Party”), the receiving party shall return all documents or other materials containing the confidential information, or destroy the same as instructed by the Disclosing Party. Upon termination of the project, the Disclosing Party shall have the right to request in writing that the receiving party return relevant confidential information.

8.2.7 The above restrictions shall not apply in the following situations:

8.2.7.1 The confidential information is legally in the possession of the receiving party at or before the signing of this Contract;

- 16 -

8.2.7.2 The confidential information has been made public or is available from the public domain at the time of notification to the receiving party;

8.2.7.3 The information is disclosed or provided to the receiving party by a third party without restriction and without breaching any confidential obligation;

8.2.7.4 The confidential information has been made public or is available from the public domain without breaching the obligations hereunder;

8.2.7.5 The confidential information has been independently developed by the receiving party or any of its affiliates at or before the signing of this Contract without benefiting from the information received from the Disclosing Party or any of its affiliates;

8.2.7.6 If the receiving party is required to disclose any confidential information by applicable laws or court order, or as required by administrative authorities (through oral questioning, inquiries, requests for information or documents, subpoenas, civil or criminal investigations or other procedures), the receiving party shall immediately notify the Disclosing Party and provide the Disclosing Party with necessary explanation of such laws, order or requirement as well as the opportunity to prevent or limit the disclosure of such information.

8.2.8 If the confidential information provided by the Disclosing Party infringes upon the intellectual property rights of a third party, the receiving party shall not be liable for such infringement and shall be exempted from any claims arising therefrom.

8.3 Period of Confidentiality

Both parties agree that the confidentiality clause hereof shall survive any change, rescission or termination of the Contract, and that each party shall continue to assume its confidentiality obligations as agreed herein regardless of such change, rescission or termination.

8.4 Default Liability

Any failure by either party to perform the confidentiality clause herein shall be deemed as a breach of contract. The party who commits the breach shall bear the losses caused to the observant party by its breach. If the observant party determines that compensation alone would be insufficient remedy for the breach hereof, it shall also be entitled to injunctive, actual performance or other reasonable remedies. Regarding the liability of compensation for losses to the observant party due to the disclosure, Party A and Party B specifically agree as follows: During the performance of the Contract, supplementary agreements shall be entered into by the parties where necessary based on the actual situation. Such supplementary agreements shall have the same legal effect as this Contract;

- 17 -

in case of any conflict between any supplementary agreement and the Contract, the terms of the supplementary agreement shall prevail.

9 Miscellaneous

9.1 Force Majeure and Exclusion of Liability

9.1.1 “Force Majeure” means any unforeseeable event beyond the reasonable control of either party (or unavoidable even if foreseeable) at the time of concluding this Contract, which prevents, affects or delays the performance by either party of all or part of its obligations hereunder. Such event includes but is not limited to, governmental action, natural disaster, war or any other similar event.

9.1.2 In case of Force Majeure events, the party becoming aware of such event shall send timely and sufficient written notice to the other, specifying the possible impact of such event on the Contract, and shall provide relevant evidence within a reasonable time.

9.1.3 If either party or both parties are unable to perform or completely perform the Contract due to Force Majeure, they shall be exempted from liability in part or in whole based on the impact of Force Majeure, except as otherwise provided by law.

9.1.4 Either party or both parties shall continue to perform the Contract within a reasonable time after the effect of Force Majeure has been eliminated.

9.1.5 Either party that delays performing this Contract before the occurrence of any Force Majeure event shall not be exempted from the corresponding liabilities.

9.1.6 During the validity of the Contract, if a government agency or chamber of commerce issues a certificate confirming that due to Force Majeure, either party is unable to continue to perform the Contract, and would be unable to do so even after Force Majeure has been removed, relevant risks, liabilities and expenses shall be equally borne by the two parties.

9.2 Effectiveness and Termination of the Contract

9.2.1 Effective Date of the Contract

9.2.1.1 This Contract is made in triplicate and shall come into force after being sealed by both parties. Each party shall hold one copy, and the third copy shall be retained by the registration authority, all of which shall have the same legal effect.

9.2.1.2 As from the effective date of this Contract, any oral or written contracts or commitments made by the parties on matters related to this Contract shall lose its legal effect.

9.2.2 Termination of the Contract

- 18 -

At the request of either party for termination, a supplementary agreement for termination of the Contract shall be entered into by the parties upon reaching agreement through friendly negotiation. This Contract shall be terminated once the supplementary agreement is confirmed by the signatures and seals of both parties.

9.2.3 Others

9.2.3.1 Any outstanding claims and debts incurred prior to expiration of the Contract shall not be affected by the expiration hereof. The debtor shall continue to fulfill its debt obligations to the creditor.

9.2.3.2 Any change to individual terms and annexes of the Contract shall be reached through friendly negotiation and made in the form of a written document executed by the authorized representatives of both parties to be valid.

9.2.3.3 To guarantee the continuity of the technical service project hereunder, the parties agree that Party B shall have the right to request Party A to renew the contract upon its expiration.

9.3 For matters not covered herein, both parties shall separately enter into a supplementary agreement which shall have the same legal effect as this Contract.

9.4 The annex to the Contract:

Annex: Technical Service Scheme Specification    

- 19 -

Service Recipient (Party A) Name Legal Representative Entrusted Agent Contact (Responsible Person) Domicile (Mailing Address) Phone Bank Name Account No. Beijing Sohu Donglin Advertising Co., Ltd. Li Wei                (signature & seal) /                 (signature & seal) Wang Meng    (signature & seal) Room 1301, 13th Floor, Building 3, Courtyard 2, Kexueyuan South Road, Haidian District, Beijing City 010-62728577 China Merchants Bank, Beijing Branch, Beisanhuan Subbranch 110907072410805 Zip Code Fax 100190 010-56412892 Special seal for contracts or common seal (seal) January 1, 2022 Service Provider (Party B) Name Legal Representative Entrusted Agent Contact (Responsible Person) Domicile (Mailing Address) Phone Bank Name Account No. Beijing Sohu New Media Information Technology Co., Ltd. Charles Zhang (signature & seal) /                 (signature & seal) Pang Li    (signature & seal) Room 1201, 12th Floor, Building 3, Courtyard 2, Kexueyuan South Road, Haidian District, Beijing City 010-56603776 China Merchants Bank, Beijing Branch, Beisanhuan Subbranch 862281851810001 Zip Code Fax 100190 010-56412878 Special seal for technical contracts or common seal (seal) January 1, 2022

- 20 -

Paste the revenue stamp here

— — — — — — —

Review & registration by the Registration Authority:
Handled By:	Technical Contract Registration Authority (special seal) Date

Annex:

Technical Service Scheme Specification

Project name:	Business Technology Application Solution Based on Big Data Analysis
Service Recipient:	Beijing Sohu Donglin Advertising Co., Ltd.
Service Provider:	Beijing Sohu New Media Information Technology Co., Ltd.

Table of Contents

CHAPTER 1 PROJECT OVERVIEW	- 1 -
1.1 TECHNICAL SERVICE BACKGROUND	- 1 -
1.1.1 Macro-environment: exploring market opportunities from long-term and short-term perspectives	- 1 -
1.1.2 Industrial status: the market is full of vitality and continues to grow	- 5 -
1.1.3 Marketing role: the digitalization wave revolutionizes the industry	- 6 -
1.1.4 Vertical industry: integration and competition create a new pattern	- 7 -
1.2 IMPLEMENTATION CONDITIONS OF THE SERVICE PROVIDER	- 12 -
1.2.1 Introduction of the enterprise implementing the project	- 12 -
1.2.2 Technical basis and resource conditions	- 13 -
CHAPTER 2 TECHNICAL SERVICE SCHEME PLANNING	- 15 -
2.1 DEVELOPMENT TREND OF NEW ONLINE MARKETING TECHNOLOGY	- 15 -
2.1.1 Technology: new marketing technology stepping into the deep-water area of data	- 15 -
2.1.2 Strategy: creative content and experience become the next trump card	- 16 -
2.1.3 Scenarios: media integration and linkage to achieve full scenario coverage	- 17 -
2.1.4 Concept: upgrading and innovation of the point-to-body mode	- 17 -
2.2 BUSINESS TECHNOLOGY APPLICATION SOLUTIONS BASED ON BIG DATA ANALYSIS	- 18 -
2.2.1 Research on online marketing technology that drives digital upgrade	- 18 -
2.2.2 Application examples of new online marketing technology	- 23 -
CHAPTER 3 PROJECT IMPLEMENTATION MANAGEMENT	- 32 -
3.1 PROJECT MANAGEMENT MODE	- 33 -
3.1.1 Definition and organization of the project	- 33 -

I

3.1.2 Plan of the Project	- 33 -
3.1.3 Tracking management of the project	- 33 -
3.2 PROJECT IMPLEMENTATION METHOD	- 34 -

II

Chapter 1 Project Overview

1.1 Technical service background

Under the wave of digitalization, the production and dissemination methods of the online advertising industry are changing at an accelerated rate, and the trend of digital marketing upgrading is becoming increasingly obvious. Post-epidemic environment also profoundly affects the transformation of enterprise marketing model. How to apply digital technology innovatively, place advertisements more accurately and efficiently to cut into the pain points of consumers, and realize the maximization of marketing value has become the most critical issue for major online advertising platforms.

1.1.1 Macro-environment: exploring market opportunities from long-term and short-term perspectives.

1.1.1.1 Short-term environment

(1) The epidemic situation of COVID-19 causing marketing instability: the epidemic has slowed down economic growth, and caused obvious impact on the advertising market.

The economic growth of China has continued to slow down in the past decade, and under the influence of COVID-19, the GDP growth rate in 2020 is less than 3%, which is the lowest level in the past 20 years. Influenced by the overall economic environment, the growth rate of advertising revenue dropped significantly after the epidemic, keeping pace with the growth rate of GDP.

Nearly 30% of advertisers think that the negative impact of the post-epidemic environment on marketing lies in “the instability of the marketing and promotion environment of offline channels”, and “the increasing amount of online information and more difficulty in screening high-quality information” and “the rapid change of consumers’ psychology and the difficulty in maintaining customer relationships” are also the influencing factors that advertisers are more concerned about this year. However, with the normalization of epidemic prevention and control, constant optimization of the internal structure of enterprises to cope with the epidemic situation, gradual recovery of the external environment and economy, and wide popularization of digital technology, advertisers begin to try and apply more diversified marketing forms, and pay more attention to the online marketing of Internet attributes, so as to realize the quantification of marketing effect and the precision of operation. The overall advertising market has also entered a stage of deep refinement and high efficiency, which is different from the previous mode of emphasizing scale and speed.

- 1 -

(2) Marketing work gradually returning to normal in the post-epidemic period: judging that the impact of the epidemic has decreased, advertisers have confidence in the marketing work According to the research report by iResearch on advertisers, over 50% of advertisers believe that the epidemic situation in 2020 had a great or very great impact on marketing. However, with the easing of the impact of the epidemic, normalization of daily prevention and control, and improvement of residents’ awareness of epidemic prevention, the impact of epidemic situation on social life is gradually weakening.

In the post-epidemic era in 2021, the proportion of advertisers who believe that the epidemic environment still has a great or very great influence on the marketing work has dropped by about 15%, while the proportion of advertisers who think that there is little or no influence has increased to about 30%. This also reflects that although the anti-epidemic work is not over yet, most advertisers consider that the post-epidemic environment is no longer the core consideration in the process of marketing work, and they show a positive and confident attitude towards the marketing environment.

1.1.1.2 Long-term environment

(1) Acceleration of social digital transformation: the epidemic has catalyzed the development of digital economy, and the digitalization of the three major industries has been comprehensively upgraded.

Digital economy is defined in G-20 Digital Economy Development and Cooperation Initiative as “a broad range of economic activities that include using digitized information and knowledge as the key factor of production, modern information networks as an important activity space, and the effective use of information and communication technology (ICT) as an important driver of productivity growth and economic structural optimization”.

In recent years, the scale of China’s digital economy has been steadily growing, with its proportion in GDP gradually increasing. Under the circumstances that the overall economic environment is impacted by the epidemic, the scale of digital economy in China has maintained a positive and good growth trend, reaching RMB39.2 trillion in 2020, ranking second in the world. At the same time, the integration of digital economy and real economy is accelerating, and digitalization penetrates into every corner of social life, promoting economic transformation and upgrading and the transformation of growth mode. The digital economy has also promoted the digital transformation of enterprises, providing new driving force and upgrading path for enterprise development.

(2) The general trend of enterprise digitalization: epidemic situation promotes digital cognition and promotes digital transformation of enterprises.

The social digital transformation also promotes the digital upgrading of enterprises and the digital transformation of marketing work, and the epidemic situation catalyzes this process.

- 2 -

After the outbreak of COVID-19, nearly 60% of advertisers believe that “the epidemic has accelerated the process of digital transformation of enterprises”. Under the epidemic environment, telecommuting has changed from “optional” to “necessary” for many enterprises. After the epidemic, more and more work such as the approval of various processes and the convening of meetings is moved to the online platform, which also improves the digital penetration rate of the company’s business processes and marketing work. In the survey on the relationship between digitalization and marketing, more than 90% of advertisers hold the opinion that that “digital transformation is the inevitable trend of the company’s marketing work”, which is a significant increase over the previous year when 67.0% of advertisers believe that digital upgrade is very necessary. One reason is that marketing, as an important business scenario of enterprises, conforms to the general trend of enterprise digitalization. The other reason is that the accelerated upgrade and penetration of marketing technology and continuous popularization of its application in different sectors and scenarios constantly increase advertisers’ confidence in digitalization of marketing work.

(3) Continuous increase of investment in marketing technology: under the joint influence of the digital wave and the epidemic, the proportion of new marketing technology has been increasing

According to the survey data by iResearch on advertisers over the years, more than 60% of advertisers invest 10% or more of their total budget in new marketing technology in the latest year. At the same time, more than 70% of advertisers have significantly increased their budget investment in new marketing technology compared to the pre-epidemic period, and about 20% of advertisers have increased the investment by more than 30%, only 20.3% of advertisers maintain the pre-epidemic level of investment and 5.9% of advertisers choose to reduce the investment in new marketing technology.

New marketing technology is an important means to promote the digital transformation of enterprises. In the process of enterprises seeking digital transformation, it is inevitable to increase the budget investment in marketing technology. At the same time, after the epidemic, the effect and value of the application of new marketing technology are constantly reflected, and advertisers’ confidence in the use of marketing technology is also continuously enhanced. With the further improvement and maturity of new marketing technology, advertisers’ investment in new marketing technology will continue to improve.

(4) Marketing taking data as the core driving force: marketing full link digital era is coming

Under the combined effects of the overall economic growth slowdown, the gradual fading of traffic dividends, the continuous penetration of digital economy and the digital transformation of enterprises, the marketing work has been transformed into more refined and intelligent digital marketing in recent years, and the epidemic situation in 2020 has accelerated enterprises’ attention to and development of digital marketing upgrade, which makes enterprises more aware of the importance of data in the process of digital marketing. Digital marketing has entered the era of data empowerment.

- 3 -

Enterprises of different industries and sizes in China have been increasing their investment in data systems. However, in the application of marketing data, it has evolved from simple data analysis’s providing reference for marketing decision-making to the precipitation based on the capitalization of user data, which can maximize the role of data in marketing, so as to move towards digitalization and quantification in terms of marketing objectives, marketing strategies, marketing ideas, marketing effects and sales volume.

(5) Data capitalization becoming core demands: data capitalization demands constant attention, and private domain management is expected to become the next flying pig

According to the research data by iResearch on advertisers, after the epidemic, more and more advertisers have turned their attention to data capitalization and private domain traffic management. In the choice of application types of new marketing technology, private domain management and customer data platform have become the first choice of advertisers, and nearly 60% of advertisers pay more attention to customer relationship management in application scenarios.

The construction of customer data platform can precipitate customers’ data assets, output effective marketing insights by mining customer data, and improve marketing accuracy and full link value. Private domain management means that after advertisers have mastered the “data assets” of consumers, including consumers’ portraits, crowd attributes, labels and other first-hand data, they can maintain customers in a “one-to-one” way that can be continuously applied, such as making phone calls and sending text messages, i.e., realizing long-term and loyal customer relations with lower cost, which has become a new opportunity that advertisers are increasingly optimistic and concerned about. For many advertisers, seizing this opportunity means the possibility of achieving “overtaking in corners” in marketing.

(6) Data security law regulating industry ecology: Data assets are recognized by legislation, and marketing digitalization has stepped into the deep-water area of industry regulation.

On June 10, 2021, the 29th Meeting of the Standing Committee of the 13th National People’s Congress passed the Data Security Law of the People’s Republic of China. On August 20, China’s first Personal Information Protection Law was passed at the 30th Meeting of the Standing Committee of the 13th National People’s Congress, which means that users’ personal information and data assets as a new and independent protection object have been recognized by legislation, and at the same time, it will have a profound impact on the digital transformation of marketing.

Under more improved laws and regulations, marketing digitalization will step into the deep-water area of industry regulation. In the short term, the industry will face certain challenges, but in the long run, all parties in the industry will pay more attention to data security and protection. It will be more compliant and orderly on the data application level, and marketing digitalization will enter the stage of high-quality development. For advertisers, it is becoming more and more difficult to rely on third-party monitoring codes for “attribution”. It is necessary to pay more attention to precipitating their own consumer data assets. How to deeply and efficiently tap their limited consumer data value within the scope of legal compliance will become the long-term marketing focus in the future.

- 4 -

1.1.2 Industrial status: the market is full of vitality and continues to grow.

1.1.2.1 Scale of China’s online advertising market

According to the survey data, the scale of China’s online advertising market reached RMB766.6 billion in 2020, with a year-on-year growth rate of 18.6%, which is 4.1% lower than last year’s estimated growth rate.

In 2020, the growth rate of China’s online advertising market slowed down significantly, mainly because some brands have re-configured and planned their online advertising budget due to the epidemic. With the continuous recovery of brand owners’ confidence in market and further improvement of business activity, it is expected that China’s online advertising market will recover to a certain extent in 2021, with a year-on-year growth rate of 21.9%. In the next three years, China’s online advertising market will continue to maintain a steady growth trend with a compound annual growth rate of 17%. The transformation and pursuit of refined, efficient and intelligent marketing by the brand owners will be the joint efforts of all parties in the industrial chain of the online advertising market, and also the core driving force for the continued growth of the online advertising market in the future.

1.1.2.2 Scale of China’s mobile advertising market

In 2020, the scale of mobile advertising market reached RMB672.5 billion, with a year-on-year growth rate of 24.2%. In 2020, the epidemic further promoted the usage habits of mobile Internet users, which made the scale of mobile advertising market still maintain a high growth, and its proportion in the overall online advertising market further increased to 87.7%.

In the next three years, the mobile advertising market will continue to develop steadily at a compound annual growth rate slightly higher than that of the overall online advertising market, which is expected to reach RMB1,174.1 billion in 2023. At the same time, as the penetration rate of mobile advertising in online advertising gradually approaches the ceiling, the growth momentum of mobile advertising in the future will come from the continuous investment of brand owners in online advertising budget and the continuous innovation of digital marketing industry.

- 5 -

1.1.2.3 Subdivision structure of online advertising market

In 2020, the share composition of different forms of online advertising in China was still continuously adjusted, among which the proportion of e-commerce advertising and information flow advertising continues to rise, ranking the top two advertising forms with 39.9% and 32.9% market share respectively. In particular, information flow advertising has become the most significant growth form, mainly because all kinds of media have begun to deepen the layout of information flow content, further increasing the commercialization space of information flow advertising.

1.1.3 Marketing role: the digitalization wave revolutionizes the industry.

(1) Advertisers’ perception of the functions of the marketing department: to gain insight into the changing trend of the market, formulate marketing strategies and promote the company’s sales.

The economic downturn keeps advertisers keen on the market environment and consumers’ psychological changes. Regarding the responsibilities of the marketing department, advertisers believe that the most important things are “understanding market changes and formulating marketing strategies” and “correctly understanding consumers and market trends”. Such data also correspond to the advertisers’ ability to improve the marketing department. On the latter issue, “the ability to gain insight into marketing trends and put forward effective strategies” and “the ability to process and analyze consumer data” rank first. On the one hand, it reflects advertisers’ great attention to insight into the market environment and consumers; on the other hand, it also shows that the ability of insight into market and analysis on consumer data, which are the core responsibilities of marketing, is still in a state to be improved in most enterprises, and there is still great room and potential for improvement.

(2) Advertisers’ planning for the overall marketing budget: the marketing budget is growing at a high speed, and the investment in network platforms and marketing technology becomes the main force

Compared with the year before the epidemic in 2019, most advertisers have increased their overall marketing budget investment in the past year, and more than 12% of them have increased their overall marketing budget by over 50%. It can be seen that advertisers still have sufficient confidence in marketing work after the epidemic. In the environment where the overall economic growth is affected by the epidemic, they still maintain the growth trend of marketing budget. Specifically, 76.6% of advertisers said that the main driving force for growth came from the marketing budget of online platforms, including search engines, portal information, social networking, short videos, e-commerce, etc. In addition, nearly half of the advertisers have increased their investment budget for marketing technology. With the continuous improvement and popularization of customer data management and marketing automation and other technologies, more and more advertisers realize the value of marketing technology.

- 6 -

(3) Advertisers’ views on digital transformation of marketing: digital transformation enhances the value of precise marketing and strengthens the linkage of various departments of enterprises.

According to the survey data, about 50% of advertisers believe that the greatest help brought by digital transformation to marketing work is “obtaining insight into customers in real time and providing more accurate marketing value”. Digitalization conforms to the trend of advertisers’ demands for accurate marketing. At the same time, nearly 70% of advertisers believe that digital transformation improves “the linkage between marketing and public relations, products and sales”. On the issue of pain points in digital transformation, the cognitive distribution of different advertisers is relatively balanced.

(4) Advertisers’ perception of the value of new marketing technologies: effect quantification and efficiency improvement are the greatest value points, and marketing efficiency is a long-term demand

According to the survey data, nearly half of the advertisers believe that the greatest value brought by the application of new marketing technology at present is “real-time tracking of marketing data” and “improving work efficiency and saving labor cost”. The demand of improving marketing efficiency is also one of the pain points that the new marketing technology is long expected to solve.

Under the background of rising marketing cost and increasing customer acquisition cost of enterprises, as new marketing technology is different from traditional advertising and event planning and has clear indicators and billing methods, advertisers hope that the application of new marketing technology can maximize the quantification of the input and output of marketing work, and then make a breakthrough in marketing efficiency.

1.1.4 Vertical industry: integration and competition create a new pattern.

1.1.4.1 Advertising market of short video industry

The growth rate of advertising on short video platforms declined, and the market scale reached RMB133.6 billion in 2020

Compared with the rapid growth in the previous two years, the growth rate of advertising revenue of short video platforms dropped to 67.1% in 2020, with a total market scale of RMB133.6 billion. From the demand side, short video advertising is still the focus of major advertisers, and the continuously optimized content ecology of the platforms continues to increase the overall number of users and user stickiness, becoming the platforms a fertile soil for advertisers’ marketing growth. On the whole, the head platforms continue to explore more commercialization possibilities, and gradually open up live advertisements and search advertisements in terms of advertising forms. It is estimated that short video advertising will keep its compound growth rate at 34.6% in the next three years to exceed RMB300 billion by 2023.

- 7 -

1.1.4.2 Social advertising market in China

In 2020, the scale of social advertising was RMB79 billion, and social marketing will help the brand continue to grow.

In 2020, the scale of China’s social advertising market was RMB79 billion, and the growth rate dropped to about 21%. It is estimated that the scale will be close to RMB125 billion by 2023. The COVID-19 epidemic in early 2020 had a great impact on the advertising revenue of various platforms in the first quarter, but the growth momentum resumed in the second to fourth quarters. Head platforms such as WeChat, Weibo, Zhihu and Xiaohongshu, which have accumulated a large number of users, integrate diverse marketing resources and methods based on their own ecological characteristics and advantages, provide differentiated social advertising marketing products to advertisers, and continuously optimize the advertising bidding system to create new increments for brands. iResearch predicts that with the joint promotion by Z-generation young users and emerging brands, the scale of China’s social advertising market will continue to grow.

1.1.4.3 E-commerce advertising industry in China

(1) The scale of e-commerce advertising market exceeded RMB300 billion in 2020, and will continue to grow steadily

E-commerce advertising is rich in forms and has a relatively direct conversion link, which provides advertisers with diversified marketing options that make it easy to realize sales conversion. At the same time, the strengthening of users’ online consumption habits has further promoted advertisers’ demand for operating online sales channels and placing e-commerce advertisements.

In 2020, the scale of e-commerce advertising market exceeded RMB300 billion, accounting for 39.9% of the total scale of online advertising, ahead of the advertising revenue of other forms of media. In 2017, the rise of social e-commerce platform represented by Pinduoduo also accelerated the growth of China’s e-commerce advertising market. In recent years, the reshuffle of social e-commerce platforms has been basically completed, and the scale of e-commerce advertising market has reached a high level. Therefore, from 2020 onwards, China’s e-commerce advertising market is expected to enter a period of steady growth.

(2) The development trend of e-commerce advertising in China: deepen the application of content-side technology, and improve the display efficiency and reach experience of e-commerce advertising

- 8 -

With respect to the combination of e-commerce platforms and application of technology in advertising, apart from the development and application of mature technology related to programmed advertising, the deepening application of content-side technologies such as programmed creative platforms and AR/VR will further help e-commerce platforms to strengthen the shopping experience of consumers and the advertising experience of brands and merchants, and improve the display and conversion efficiency of e-commerce advertising. At present, head e-commerce platforms such as Taobao Tmall and JD.COM have joined in the exploration and application of such technologies. With the further maturity of various technologies such as AI, AR and VR, it is expected that more e-commerce platforms will continuously improve the display efficiency and reach experience of platform advertising with the help of the above technologies.

1.1.4.4 Online video industry advertising market in China

(1) In 2020, the market scale was RMB33 billion, a year-on-year decrease of 10.2%

Affected by the epidemic situation, advertisers’ overall investment budget and confidence declined. At the same time, during the severe epidemic situation at the beginning of the year, some market actions of advertisers were suspended due to the suspension of production. Although they were released and warmed up with the improvement of the macro environment in the second half of the year, the overall situation still showed a downward trend. The overall trend of the advertising industry was reflected in the online video industry, showing a decline in advertising revenue.

On the other hand, from the overall situation of the Internet pan-entertainment market, on the premise that the online advertising market is stabilizing, the relatively strong trend of some other types of pan-entertainment services also has a certain impact on the budget allocated by advertisers to the online video industry.

In addition, from the perspective of business logic within the industry, the core revenue source of online video industry has shifted from advertising to content payment, and the scale of paid members has further expanded, which affects the overall exposure flow of pre-rolls. In the short-term structural adjustment, there is a certain trade-off between the two.

(2) The development trend of online video industry in China:

- 9 -

Industrialized development throughout the whole industrial chain promotes standardization and improves the overall operation efficiency

For a long time, video content production in China has been in a non-standardized “team system” state, making the overall controllability of production management weak, the rights and responsibilities of each subdivision unclear, the overall budget allocation unbalanced, links separated, and the accumulation of experience difficult to remain at the enterprise level. The cost of long video content is high, and the controllability and efficiency of content cashability are low because of the non-standard industry. With the diversification and vertical development of video content, new categories and forms also raise demands and requirements for the industrial production of content. These newly emerging large-scale categories in the digital and Internet era are also less restricted by the old system because of their “newness”, which makes it easier to explore and develop cooperation modes among different aspects that support the industrial production system, thus making up for the shortcomings of the entire video content industry, improving the business efficiency of enterprises and evolving towards standardized production and integrated management.

Based on verticalization and industrialization, explore the popularization of vertical categories and improve the efficiency of content cashability

The low efficiency of cashability and controllability of video content is a problem that long plagues the industry. With the expansion of scale of audience connected by video platforms, the continuous enhancement of industrial chain’s control, and the effective exploration of content payment mode as a whole, the idea of 2C-based distribution in video content industry has been determined, and with the continuous development of users’ overall content payment habits, a new driving force is brought to the industry to improve the efficiency of content cashability. The continuous deepening of accurate audience positioning and systematic content quality control are helpful for further enhancing individual users’ willingness to pay for content. At the same time, the development of industrial production level will continuously improve the stable supply and cost control of content, which will jointly promote the branding of content categories and content theaters, continuously accumulate circle audiences, make the content develop towards vertical popularization, and then provide the cornerstone for the birth of a more advanced distribution mode.

1.1.4.5 News and information industry advertising market in China

(1) News and information industry advertising market scale in China

In 2020, the scale reached RMB64.9 billion, benefiting from the boost of demand and attention under the macro influence

In the 2nd quarter of 2020, the scale of Internet news and information advertising reached RMB64.9 billion, a year-on-year increase of 11.9%. The complicated social situation with COVID-19 epidemic as the leading factor has driven people’s demand and attention for news and information, and also promoted advertisers’ placement level on news and information platforms. Although the overall budget, confidence and rhythm of advertisers declined during the severe epidemic situation at the beginning of the year, the follow-up recovery was relatively rapid.

- 10 -

In the long run, the concentration of the news and information industry is expected to continue to increase. With limited space for user penetration and advertising load increment, it is expected that the growth rate of Internet news and information advertising scale will slow down as a whole.

In 2020, the scale of mobile news and information advertising reached RMB58.28 billion, accounting for nearly 90%

In 2020, the scale of mobile news and information advertising reached RMB58.28 billion, a year-on-year growth rate of 12.9%, accounting for 89.8% of the overall portal and information advertising market scale. Head participants in the industry continue to take information content as the origin, and gradually become a comprehensive information platform integrating services in various content forms. On the one hand, the breadth, depth and presentation forms of news and information content are continuously enriched, and with the overall trend of normalization of video carrying of information, the trend of short video information will be more obvious; on the other hand, it also extends horizontally to social services, MCN-like generation operation services for vertical experts and Internet celebrities, etc., consolidating moat and promoting mobile news at the same time.

(2) The development trend of news and information industry in China: the dimension of information bearing and transmission is gradually improved, and new content forms meet new needs

With the popularization of China Mobile Net, the intellectualization of mobile phones and the fragmentation of time brought about by the change of people’s lifestyle, the development of technology and the upgrading of users’ consumption demand make the news and information content acquisition channels show a trend of migration and penetration from offline to PC and then to mobile terminal. At the same time, the content forms gradually develop from text to pictures to video, and the transmission rate, richness and density of information are advanced from low dimension to high dimension.

- 11 -

1.2 Implementation conditions of the Service Provider

1.2.1 Introduction of the enterprise implementing the project

Beijing Sohu New Media Information Technology Co., Ltd. was established in June 2006 with a registered capital of USD60 million. Its business scope includes: development of Internet advertising technology, advertising technology service and consultation; design, production, release, and agency of various domestic and foreign advertisements

SOHU.COM is a leading Internet media, entertainment and online game group in China. It owns the Nasdaq-listed company Sohu (NASDAQ: SOHU) and Changyou, an online game developer and operator. It is one of the leading Internet brands in the Chinese world. Sohu is a well-known brand in China, and it is also the sponsor of Internet content service for 2008 Beijing Olympic Games. Sohu provides comprehensive network services for more than 700 million Internet and mobile Internet users in China.

On November 1, 1995, Dr. Zhang Chaoyang returned to China from Massachusetts Institute of Technology. In August of the following year, based on venture capital, he founded Aitexin Information Technology Co., Ltd, predecessor of Sohu. In February 1998, Aitexin launched Sohu, the first large-scale classified query search engine in China, and the Sohu brand was born. “Going out by map and surfing the Internet by Sohu”, Sohu has thus opened the magic door for Chinese netizens to access the Internet world.

In 1999, Sohu launched the news and content channel, which laid the embryonic form of comprehensive portal website and opened the era of Internet portal in China. Because of his outstanding contribution to the spread and commercial practice of the Internet in China, Mr. Zhang Chaoyang was rated as one of the “50 digital heroes in the world” by Time of the United States, and was invited to Fortune forum and appeared on the cover of Asia Week.

On July 12, 2000, Sohu was officially listed on NASDAQ (NASDAQ:SOHU), developing from a well-known domestic enterprise to an international brand. In the same year, Sohu acquired ChinaRen, a leading youth community in China, and established its position as a leading Chinese website in China. In the third quarter of 2002, Sohu achieved full profit in the domestic Internet industry for the first time, which was an epoch-making milestone in the development of China’s Internet, and promoted the full popularity of Chinese concept stocks on NASDAQ. In November 2005, Sohu signed a contract to become the sponsor of Internet content service for 2008 Beijing Olympic Games. On April 2, 2009, Changyou, a subsidiary of Sohu, was successfully listed on NASDAQ. As a result, Sohu became the first Gemini of Chinese Internet companies on NASDAQ. In September 2009, Sohu Video launched China Online Video Anti-Piracy Alliance, which played a decisive role in promoting the copyright of online videos. On April 24, 2013, the number of users of Sohu News Client, a mobile Internet product of Sohu, exceeded 100 million, making it the first news client with over 100 million users in China. In November 2017, Sogou, a subsidiary of Sohu, was officially listed on New York Stock Exchange. In May 2018, the listed holding parent company of Sohu Group moved to be registered on the Cayman Islands. In 2020, Sohu News Client covered 700 million users. In April 2020, Changyou was privatized and delisted from NASDAQ, becoming a wholly-owned subsidiary of Sohu. In September 2021, Sogou was privatized and became a wholly-owned subsidiary of Tencent.

- 12 -

As the best Internet brand in China, Sohu has attracted the attention of, joined hands with and cooperated with many internationally renowned brands, opening up an unprecedented wonderful space for Chinese netizens. At present, as the exclusive partner of Disney’s official website, NBA Chinese official website, Yao Ming personal official website, Miss World Chinese official website, and sponsor of 2008 Beijing Olympic Games Internet content service, etc., Sohu has become the best channel for Chinese Internet users to access the Internet.

Sohu has always been committed to building a century-old Internet brand and becoming a lasting, successful and great Internet company. With all-round high-quality content services and advanced Internet products, Sohu provides the most reliable online life platform for the majority of netizens, making Sohu an indispensable part of the life of the Chinese people.

Up to now, Sohu has been committed to meeting users’ Internet needs and experiences, and has grown into a leader of China’s Internet. Sohu has influenced 80% of Internet users in China, becoming their mainstream information media and the platform of life, entertainment, communication and interaction.

1.2.2 Technical basis and resource conditions

With its strong competitive strength, Sohu has developed into a super Internet platform with many well-known products, including: media (SOHU.COM, Sohu News Client, Sohu on mobile phone, Sohu Information Client, Sohu Focus), video (Sohu Video, Sohu Video Client), social networking (Huyou APP) and games (Dragon Oath series games, 17173 Platform).

As a leading Internet brand in China, Sohu has built a unique ecology of media, entertainment, socialization and marketing, which can not only customize the overall solution of multi-form media advertising, but also continuously innovate and breakthrough in network products and marketing means, continuously create diversified advertising and marketing ideas, enhance the interaction between audiences and brands, and boost the brand value of advertisers.

- 13 -

In terms of brand advertising, based on various product forms of resource platforms, media matrix, vertical channels and mobile services, Sohu has all kinds of advertising products including banner advertising, text link advertising, button advertising, video advertising, multi-type rich media advertising and specific sponsorship advertising on the website. Therefore, Sohu can not only provide daily charge service for various advertising forms, but also provide fixed payment service during the contract period; and can also provide CPM and other effect payment methods. At the same time, Sohu also provides a variety of advertising preferential choices for advertisers with long-term cooperation relationship to ensure the maximum benefit of advertisers.

In addition, Sohu has been constantly making breakthroughs in marketing innovation. On the one hand, Sohu put forward the strategy of “activity being content”, and created a number of marketing activities in vertical areas, providing advertisers with opportunities to break the circle with the influence effect of activities, and enhancing brand popularity among users with the strong output of high-quality advertising content. On the other hand, driven by technology, Sohu has continuously iterated on the video live broadcast technology, which is widely used in various content marketing activities. At the same time, it integrates value live streaming and social elements to form a live-streaming e-commerce in the style of variety show. Combined with innovative media forms such as program IP, column IP and activity IP, Sohu distributes brand advertisements through multiple scenarios and forms, helping advertisers to further expand the brand effect on the basis of accurately reaching the target audience.

In terms of talent building, since its establishment, the company has gathered a number of R&D team members with doctoral and master’s degrees and research fellows and senior engineers. At the same time, in order to attract more senior talents and create a better research environment, the company has invested heavily in purchasing servers and other hardware equipment, and cooperated with the mature R&D management system and operation and maintenance guarantee system of the group company to ensure the safety and stability of product services, which fully reflects the company’s scientific and technological development and transformation capabilities in the Internet field, and the company is therefore recognized as a “National High-tech Enterprise”. On the other hand, in order to promote scientific research projects as soon as possible and make full use of the advantages of scientific and technological talents of higher learning institutions, the company actively cooperates with famous universities and research institutions to carry out industry-university-research activities.

- 14 -

Chapter 2 Technical Service Scheme Planning

2.1 Development trend of new online marketing technology

2.1.1 Technology: new marketing technology stepping into the deep-water area of data

With the continuous development and improvement of data technology, strengthening the management of customer data and realizing the refined operation of marketing have become the type of new marketing technology that advertisers are most concerned about in the future. Private domain operation is favored by many advertisers because of its characteristic value.

- 15 -

For advertisers, the marketing demands that they most hope to solve through new marketing technologies are “intelligence and transparency of advertising”, “automatic import and calculation of traffic, construction of private domain operation matrix” and “connecting with other departments’ digital business to improve work efficiency”. At present, the application of new online marketing technology is still in its infancy, and with the in-depth penetration of technologies, it will enter the era of comprehensive empowerment of new technologies.

2.1.2 Strategy: creative content and experience become the next trump card

Advertisers’ preference shows pendulum effect, and content marketing becomes the strategic focus in the future

The development of the Internet has led to the continuous emergence of new economic forms, and has also brought about huge changes in the mode of advertising and marketing. Advertisers’ preference has shifted from traditional 4A creativity and production to traffic thinking of refined operation, and more attention is paid to the accuracy and quantitative effect of advertising.

- 16 -

With the iteration and upgrading of advertising technology in the future, after advertisers have satisfied their effect-oriented marketing demands, they will largely turn to advertising forms that are highly integrated with technology and content. The content marketing strategy, which has gradually developed recently, has attracted more and more attention from advertisers because it carries more in-depth and rich marketing information, and is easier to be combined with the content to generate stronger emotional resonance. In terms of advertisers’ optimism, social platforms and short video platforms, as the media most closely combined with content, are favored by most advertisers.

2.1.3 Scenarios: media integration and linkage to achieve full scenario coverage

Digitalization runs through the entire marketing process to boost marketing upgrades

With the deep penetration of digitalization into various industries, the increasing diversification of online and offline channels, the escalation of consumers’ demand and expansion of retail scenarios, it is increasingly difficult for a single medium to meet the marketing demands of advertisers and information demands of consumers in the environment of information overload in the future. Therefore, strengthening the Internet attributes of various media, breaking through the barriers between different media, and realizing the comprehensive coverage of consumption scenarios and the in-depth interaction of user systems through the media matrix will become the mainstream of future marketing development.

In the field of digital marketing in the future, advertisers expect to establish a perfect and rich omni-channel system, realize the connection between online and offline contacts and consumers through intelligent decision-making, and implement targeted marketing strategies.

2.1.4 Concept: upgrading and innovation of the point-to-body mode

Single tool empowerment to business line optimization, cross-departmental platform to marketing mode transformation

Under the general trend of digitalization, the digital transformation of enterprises is accelerating and deepening. As far as the current development is concerned, its transformation process can be divided into four stages: point-line-surface-body. The digital concept is from the application of single digital technology to the optimization of the whole business process, then to the establishment of cross-departmental digital platform, finally breaking the boundary of traditional business concept and realizing the innovation and change of marketing mode of enterprises.

- 17 -

Marketing, as the core business department, will further strengthen the linkage with brands, products, finance and other departments in the process of digital upgrading. The traditional single business sector division of labor will gradually disappear, and replaced by the operation thinking of multi-department linkage and coordinated development. The traditional marketing department has been upgraded to digital-driven marketing team, channel team, customer operation team, etc., and the marketing technology runs through the whole link of the marketing process through data collection, calculation, application and return in each link. At the same time, the problems existing in the traditional organizational structure, such as fragmented business segments, unclear powers and responsibilities, and high communication cost, will also be mitigated with the deepening and transformation of the digital concept, and will be upgraded in the direction of institutionalization of business processes, quantifiable and attributable effects, clear KPI assessment and so on.

2.2 Business technology application solutions based on big data analysis

2.2.1 Research on online marketing technology that drives digital upgrade

2.2.1.1 User gender prediction method, device, equipment and storage medium

(1) Background technology

This research relates to the technical field of data processing, more specifically, to user gender prediction method, device, equipment and storage medium.

In the daily operation of the network platforms, there are many scenarios that require targeted advertising based on the gender of the users, for example, advertising cosmetics for female users and advertising technology for male users.

In the prior art, targeted advertising is usually carried out according to the gender information filled in when users register, but such information is not necessarily true. If targeted advertising is carried out according to the gender information filled in when users register, the accuracy of advertising will be low and the advertising effect will be poor due to the inaccuracy of users’ gender. Moreover, users do not necessarily fill in gender information when registering, or they may fill in gender information as confidential, thus the lack of gender information will also lead to low accuracy and poor advertising effect. It can be seen that there exists the problem that the gender of online users cannot be accurately identified.

- 18 -

Therefore, there is an urgent need for an effective user gender prediction scheme to accurately identify the user’s gender.

(2) Summary of technical research contents

According to the user gender prediction scheme in this technology, the user characteristic data of the target user on the network platform is obtained, and the pre-trained user gender classification model is used to predict the gender attribute of the target user according to the user characteristic data. The user gender classification model is obtained by constructing a training sample with the characteristic data of sample users with gender labels and training the neural network model with the training sample. Based on the user characteristic data and the user gender classification model, this technical scheme can accurately predict and output the gender attributes of the target users, instead of just relying on the gender information filled in by the target users for gender identification, thus providing sufficient support for accurate targeted advertising.

2.2.1.2 A calculation method and system of attention

(1) Background technology

This research relates to the technical field of data processing, more specifically, to a calculation method and system of attention.

With the development of the Internet, users will generate a large amount of reading data when reading news, periodicals and other articles. By analyzing the large amount of reading data, we can get users’ attention to specific content in specific industries.

At present, the way to analyze reading data is to extract keywords from articles read by users, and to determine users’ attention to specific content in specified industries according to the frequency of each keyword. However, the keyword of a certain industry may appear in articles of different fields, and its influence in different fields is different. Therefore, the current way of analyzing reading data can’t accurately calculate the attention of users to specific content of the specified industry.

(2) Summary of technical research contents

The method and system for calculating the degree of attention described in this technology is: to obtain several articles containing the keywords to be analyzed, and obtain the corresponding author and reading times of each article; to obtain the weight of the keywords to be analyzed in each article in the preset keywords library; to obtain the influence degree of the author corresponding to each article; to use the weight of the keywords to be analyzed in each article, the influence degree of the corresponding author of each article and the reading times of each article to calculate the user’s attention to the keywords to be analyzed. In this scheme, the user’s attention to the keywords to be analyzed is calculated by using the reading times of each article containing the keywords to be analyzed and the influence degree of the author, and the weight of the keywords to be analyzed in each article. By using the word frequency of the keywords to be analyzed, the weight of the keywords to be analyzed in different articles and the influence of the authors of the articles, the accuracy of calculating the attention of users to the keywords to be analyzed is improved.

- 19 -

2.2.1.3 A data processing method, device and electronic equipment

(1) Background technology

This research relates to the field of data processing, more specifically, to a data processing method, device and electronic equipment.

When using queues to process tasks, the traditional queue mechanism is FIFO (First-In-First-Out). The defect is that the scenarios where priority shall be given to important tasks cannot be realized.

However, in the advertising business scenario of big data, a large number of tasks will be generated at the same time, and it is hoped that important tasks can be given priority in getting resources and be executed preferentially.

(2) Summary of technical research contents

According to the data processing method, device and electronic equipment described in this technology, when executing tasks, the tasks to be executed will be sorted according to the task priority to obtain a sorting result, and the tasks to be executed will be stacked in turn according to the sorting result to obtain a queue to be processed, and the tasks to be executed will be executed in turn according to the order of the tasks to be executed in the queue to be processed. As in this technical scheme tasks are processed according to the task priority, important tasks can be given priority in getting resources and be executed preferentially.

2.2.1.4 A method and device for predicting traffic

(1) Background technology

This research relates to data processing technology, more specifically, to a traffic prediction method and device.

Advertisers get accurate target crowd traffic by setting targeting conditions. The finer the targeting conditions, the more accurate the traffic they get, but the smaller the crowd coverage, the less the advertising exposure will be. Advertisers need to find a certain balance between obtaining accurate target crowd traffic and ensuring certain exposure. Therefore, advertisers set some combinations of targeting conditions to estimate the amount of traffic that these combinations of targeting conditions can obtain.

- 20 -

(2) Summary of technical research contents

According to the traffic prediction method and device described in this technology, a traffic matching library is first established, and the traffic matching library includes the corresponding relationship between labels and traffic. The corresponding relationship between labels and traffic is encoded with the first encoding method and stored, and then a traffic estimation request is obtained. The traffic estimation request includes label data, and the label data are encoded with the first encoding method. Use bit operation to match the encoded label data with the labels in the traffic matching library to obtain a matching result, and the estimated traffic of the label data is determined according to the matching result and the corresponding relationship between labels and traffic. According to the traffic estimation method and device, the corresponding relationship between labels and the traffic in the traffic matching library and the label data received are processed with binary coding, so that the matching efficiency can be greatly improved by bit operation in subsequent label matching to realize the rapid traffic estimation of the determined label data.

2.2.1.5 A method and system for determining information similarity

(1) Background technology

This research relates to the technical field of data processing, more specifically, to a method and system for determining information similarity.

In the field of information recommendation, the commonly used information recommendation algorithm is collaborative filtering algorithm. When using collaborative filtering algorithm for information recommendation, it is necessary to calculate the similarity between information.

At present, the way of calculating the similarity between information with collaborative filtering algorithm is to calculate the similarity between information according to the co-occurrence times of information. However, in practice, users will produce a large number of sparse behaviors, and there is too little co-occurrence relationship between information, which will lead to less information covered by collaborative filtering algorithm, and then lead to lower accuracy of information recommendation.

- 21 -

(2) Summary of technical research contents

The method and system for determining information similarity described in this technology is that: based on the time when information is consumed, sort the information consumed by each user to obtain the corresponding first information behavior sequence; integrate the first information behavior sequences corresponding to all users to obtain second information behavior sequence; calculate the first similarity between every two pieces of information with co-occurrence relationship in the second information behavior sequence; based on the second information behavior sequence, construct the topological graph between information; calculate the second similarity between each pair of nodes in the topological graph with the graph convolution algorithm; based on the first similarity between every two pieces of information with co-occurrence relationship in the second information behavior sequence and the second similarity between each pair of nodes in the topological graph, determine the final similarity between every two pieces of information in the second information behavior sequence. On the basis of calculating the first similarity between information with co-occurrence relationship, calculate the second similarity between any two pieces of information with or without co-occurrence relationship with graph convolution algorithm and the constructed topological graph, and the final similarity between any two pieces of information can be obtained by combining the first similarity and the second similarity, so as to increase the amount of information covered by the way of calculating the similarity between information and further improve the accuracy of information recommendation.

2.2.1.6 Feature generation method, device, electronic device and computer storage medium

(1) Background technology

This research relates to the field of computer technology, more specifically, to feature generation method, device, electronic equipment and computer storage medium.

At present, in the process of using machine learning model, work is carried out with the complicated feature engineering. Feature engineering refers to a series of engineering processing on the original data, which are refined into features and used as input for algorithms and models. Essentially, feature engineering is a process of representing and presenting data.

However, in the prior art, the process of extracting features can only be manually extracted by algorithm engineers, which consumes a lot of energy and time of the algorithm engineers.

- 22 -

(2) Summary of technical research contents

According to the feature generation method, device, electronic equipment, and computer storage medium described in this technology, the feature generation method includes: firstly, acquiring target configuration information; the said target configuration information includes at least one table name and at least one field name; then, aggregate the table name, the field name and the feature name to obtain at least one first configuration information; then, generate a structured query statement in a format corresponding to the data warehouse for each of the first configuration information; then, execute the structured query statement, obtain the query result, and read the field identifier of each row in the query result; for the identification of each field, determine the feature generation mode of the feature name corresponding to the field; finally, at least one first feature corresponding to the field is generated according to the feature generation mode. Therefore, the purpose of generating features quickly is achieved, and algorithm engineer is no longer required to refine features manually.

2.2.2 Application examples of new online marketing technology

2.2.2.1 An advertisement recommendation method and system

(1) Technical application background

This technology relates to the field of control, in particular, to an advertisement recommendation method and system.

At present, advertising recommendation usually adopts the random recommendation method. For example, when watching a video through a webpage, the webpage usually plays an advertisement before playing the video, and the advertisement is randomly played, but some users are not interested in these advertisements, which results in that although there are advertisements playing, there is no positive effect on promotion of the product in the advertisement.

(2) Key points of technical application

According to the advertisement recommendation method and system described in this application, the first product corresponding to the first advertisement and the first user set of the first product are determined, and at least one related datum of the first user set is determined. Based on at least one related datum, model training is carried out on the data to be trained through at least two models, and the training results of the data to be trained for at least two models are determined. Based on the training results of at least two models of data to be trained, the similarity between the user to be trained and the first set of users that at least includes the data to be trained is determined. Based on the similarity between the user to be trained and the first set of users, determine whether to recommend the first advertisement for the user to be trained containing at least the data to be trained. In this scheme, firstly, the product of the first advertisement and the first user set of the product are determined, and then, based on the related data of the first user set, the users to be trained who meet the similarity requirements with the related data of the first user set and at least contain the data to be trained are determined, so that the advertisement of the product is recommended for similar users based on the audience of the product, and the targeted recommendation is realized, and the promotion effect of the advertisement is improved. Moreover, in this scheme, the similarity is determined by training various models, which improves the accuracy of similarity determination.

- 23 -

2.2.2.2 A method, device and server for advertising

(1) Technical application background

This technology relates to the technical field of big data, more specifically, to an advertising method, device and server.

Internet advertisement is the advertisement that advertisers put on the Internet by means of advertising banners, text links, multimedia, etc. through advertising platforms. The advertising platforms will feed back advertising data to advertisers, such as display volume, click volume, conversion volume, advertising unit price and so on.

However, at this stage, the customer budget will be consumed too quickly or too slowly in the advertising time, and cannot cover a wider audience.

(2) Key points of technical application

According to the advertising method, device and server described in this application, the matched historical loading log of the target advertisement group can be obtained to determine the estimated consumption curve of the target advertisement group on that day, so as to adjust the estimated consumption rate at the current moment by comparing the current estimated consumption on the estimated consumption curve with the current actual consumption, and control the advertising according to the estimated consumption rate at the current moment, so as to realize the uniform consumption of customer budget during the advertising. This can make advertisers’ customers cover a wider audience and improve the conversion rate.

- 24 -

2.2.2.3 A method and device for allocating advertising budget

(1) Technical application background

When the advertiser sets the budget, the system will automatically launch the advertisement. If the advertiser’s bid is relatively high, it will run out quickly in a short time. Therefore, for the advertiser:

a. budget consumption is too fast for the advertiser to participate in later biddings;

b. it is not possible to compare the conversion effects during different time periods;

c. it is not possible to contact different customers;

d. it will have a negative impact on the conversion effect.

For the platform:

a. competition is too concentrated, which affects the stability of the system;

b. it is not possible to provide advertisers with more options to optimize the effect.

This technical application aims to make the advertising group budget spent more reasonably in the time dimension, so that the budget consumption is directly proportional to the traffic consumption.

(2) Key points of technical application

In order to solve the problems existing in the application of the prior art, the big data real-time streaming scheme can be used for development, but the technical difficulty, the maintenance cost, and the latency are high, especially when there are many advertisement groups, huge machine resources are occupied and the cost performance is extremely low.

Therefore, the application of this technology is to process all curve data and evaluate the authenticity of the curve through big data technology according to the real data of the company, then sort out and analyze the historical log data, estimate the traffic curve data of the next day, then allocate the budget of the advertising group to each minute according to the traffic proportion, and then track the consumption every minute to adjust the consumption rate, so as to achieve the purpose of uniform consumption, and can dynamically modify the uniform consumption curve according to the orientation conditions modified by advertisers to dynamically adjust the budget consumption. In addition, the remote call scheme can be used without affecting the current program, and the capacity can be dynamically expanded to respond to the impact of large traffic on the system.

- 25 -

2.2.2.4 Method and device for product recommendation

(1) Technical application background

This technology relates to the technical field of data mining, in particular, to a method and device for product recommendation.

With the popularity of the Internet, more and more users begin to use online shopping platform (referred to as “OSP”). Users can browse the information of products (including virtual products and physical products) on the web pages provided by OSP, and then select and consume some of the products.

In order to select products conveniently, OSP generally provides product recommendation function, that is, using recommendation algorithm to predict the products that users prefer, and recommending the products that users prefer when they visit OSP.

Collaborative filtering algorithm is a kind of existing recommendation algorithm. Its principle is that for every two users, the real behavior sequences of these two users are obtained (the real behavior sequences are used to record the products consumed by the users in a certain period of time), and the user similarity of these two users is determined based on the co-occurrence times of these two real behavior sequences (the times that the same product appears in two real behavior sequences), and then some products consumed by users are recommended to other users with high user similarity.

The accuracy of the collaborative filtering algorithm depends on the length of the real behavior sequence used for calculation (referring to the number of products recorded by the real behavior sequence). For many users with short real behavior sequences, the accuracy of the user similarity among these users calculated through the collaborative filtering algorithm is low, so that accurate product recommendation cannot be carried out.

(2) Key points of technical application

According to the method and device for product recommendation described in this application, the real behavior sequences of multiple candidate users are obtained; the nodes corresponding to the products of the real behavior sequence and the edges connecting the nodes are used to construct the topological graph of the product; the weight of edges is determined according to the product similarity of products connected by the edges; random walk is performed on topological graph of the product to obtain the topological behavior sequence of each candidate user, and the topological behavior sequence and the real behavior sequence are spliced to obtain the supplementary behavior sequence of the candidate user; the topological behavior sequence of the candidate users includes the products corresponding to each node visited during random walk; user similarity among candidate users is calculated with the supplementary behavior sequence of candidate users, and the user similarity among candidate users is used for product recommendation. By splicing the topological behavior sequence and real behavior sequence obtained by random walk, this scheme can increase the length and co-occurrence times of the behavior sequence of candidate users, thus improving the accuracy of product recommendation.

- 26 -

2.2.2.5 Video recommendation method and device, storage medium and electronic equipment

(1) Technical application background

This technology relates to the technical field of data processing, in particular, to a kind of video recommendation method and device, storage medium and electronic equipment.

In recent years, with the rapid development of Internet technology, users of various video platforms are also increasing. All kinds of videos contain a lot of rich and interesting content, so watching videos has become an important entertainment activity in people’s daily life. However, with the increasing number of videos, it is difficult for users to quickly get the videos they are interested in from the massive video resources.

In the prior art, in order to enable users to get the videos they are interested in from a large number of videos, videos similar to those watched by users are usually recommended by identifying video titles, video key frames or audio of videos. However, if the user does not have a video viewing record, the video that the user is interested in cannot be perceived.

(2) Key points of technical application

This application relates to a video recommendation method and device. The method comprises the following steps: when receiving a video recommendation request for a target user, generate feature information according to basic information of the target user, wherein the target user is a user who does not have a video viewing record in a preset time period; call a preset classification model to determine the user type to which the target user belongs based on the feature information; determine a pre-established video set to be recommended corresponding to the user type; wherein, the video set to be recommended contains a plurality of user-preferred videos; the user-preferred videos are videos that meet preset preference conditions among watched videos of each historical user belonging to the user type; recommend the user-preferred videos in the video set to be recommended to the target user. It can recommend videos that users are interested in when there is no video viewing record of the users.

- 27 -

2.2.2.6 A video recommendation method, target service provider, service caller and system

(1) Technical application background

This technology relates to the technical field of video recommendation, more specifically, to a kind of video recommendation method, target service provider, service caller and system.

With the development of the Internet, more and more video software appears in the application market. When applying video software, the corresponding video will be recommended to users according to user features and video features.

At present, the way of video recommendation is: combining user features and video features, using machine learning model to predict the click rate of video, and recommending videos according to the predicted click rate. However, as the number of videos increases and the length of video features becomes longer, it takes a longer time to process videos when recommending videos to users, and a large number of videos cannot be processed due to the low ability to process videos and the high latency of recommending videos.

(2) Key points of technical application

This application describes a video recommendation method, target service provider, service caller and system. The method includes: the service caller obtains the server instance list and determines the target service provider from the server instance list according to the preset load balancing strategy. The service caller sends the user ID of the user to be processed and the video ID of a plurality of videos to be recommended to the target service provider. The target service provider obtains the user features according to the user ID and the video features according to the video ID, processes the user features and multiple video features by using the preset video recommendation model to obtain the predicted click rate of multiple videos to be recommended, and feeds back the predicted click rate of multiple videos to be recommended to the service caller. By setting up multiple service providers and selecting the corresponding target service providers for video recommendation by using load balancing strategy, the video processing ability is improved, thus reducing the latency of video recommendation.

2.2.2.7 An information recommendation method and device

(1) Technical application background

- 28 -

This technology relates to the technical field of information recommendation, more specifically, to an information recommendation method and device.

In the practical application of information recommendation, DeepFM model is usually used as the click rate prediction model to predict the click rate of information.

In the process of using the DeepFM model to predict the click probability of information, in order to improve the prediction accuracy of the click probability, it is necessary to increase the dimension of the DeepFM model. This way will lead to the fact that the DeepFM model consumes a large amount of computing resources, and it will take a long time to predict the click probability of information, i.e., predicting the click probability takes up a large amount of computing resources and is high in latency.

(2) Key points of technical application

The information recommendation method and device of this application comprises the following steps: receiving a recommendation request at least containing user features sent by a target user; acquiring historical operation data of target users, and recalling information in the information base according to historical operation data and user features to obtain multiple pieces of target information of different categories; for each piece of target information, the article features, user features and context features of the target information are spliced to obtain the spliced features corresponding to the target information; for each piece of target information, obtain the cache vector of the splicing feature corresponding to the target information from the cache vector of the splicing feature corresponding to each piece of information in all the pre-determined and cached information bases; inputting the splicing features and cache vectors corresponding to all target information into the recommendation model for click rate prediction, and obtaining the predicted click probability of each piece of target information; feeding back the target information whose predicted click probability is greater than or equal to the click rate threshold to the target user. By predetermining and caching the cache vector of the splicing feature corresponding to each piece of information in the information base, when the DeepFM model is used to predict the click rate, the corresponding cache vector is directly searched from the cache to predict the click rate, so that the prediction accuracy of the DeepFM model is guaranteed, and the computational amount, the occupied computing resources, and the latency of predicting the click rate are reduced.

- 29 -

2.2.2.8 Item recommendation method and device, electronic equipment and computer storage medium

(1) Technical application background

This technology relates to the technical field of item recommendation, in particular, to an item recommendation method and device, electronic equipment and computer storage medium.

In order to promote and sell articles, videos, commodities and other items, and to better meet the demands of users, nowadays, items are recommended based on the recommendation algorithm.

Nowadays, the recommendation method is mainly based on collaborative filtering algorithm. Its principle is to discover the preference of users based on mining the historical behavior data of users, and predict the items that users may like for recommendation. According to the specific co-occurrence times between items, that is, based on the times of interaction between items and users, such as the times of purchase, favors or likes, the similarity between items is calculated, and then items are recommended according to the similarity between items.

However, this method depends on the historical data of users’ interaction behavior, so it can only be used for items with co-occurrence. When there is no co-occurrence between two items, it will be impossible to calculate the similarity between items, thus impossible to recommend items based on the similarity between items.

(2) Key points of technical application

Based on the above-mentioned shortcomings of the prior art, this technical application provides an item recommendation method and device, electronic equipment and computer storage medium, in order to solve the problem that the similarity can’t be calculated for the existing items that do not have the co-occurrence of interaction behavior with users.

The item recommendation method described in this application is used to count the interaction set between the user and the item, and the association set between the item and the item attribute. The interaction set includes the set of items that each user interacts with and the set of users that each item interacts with. Association set includes the set of item attributes contained in each item, and the set of items associated with each item attribute. Then, based on the interaction set, the number of items interacted by every two common users of every two items is calculated to obtain a plurality of first co-occurrence quantities corresponding to every two items; and based on the association set, the number of items associated with every two common item attributes corresponding to every two items is calculated to obtain a plurality of second co-occurrence quantities corresponding to every two items. Among them, the common users of two items refer to users who interact with both items. The common item attribute corresponding to two items refers to the item attribute contained in both items. Finally, the first co-occurrence quantity and the second co-occurrence quantity corresponding to each two items are respectively used to calculate the similarity of every two items, so that the similarity can be calculated by the co-occurrence times of the items in user behavior and the co-occurrence times of the items in item attributes, and as the items all have corresponding item attributes and can well reflect the similarity among the items, the similarity between items can be accurately calculated even if there is no co-occurrence of users’ behaviors among items, and then it can be ensured that items can be recommended to users based on the similarity of every two items.

- 30 -

2.2.2.9 Multimedia information recommendation method, related device and computer storage medium

(1) Technical application background

This technology relates to the technical field of computers, in particular to a multimedia information recommendation method, related device and computer storage medium.

Internet has become the main source for people to obtain multimedia information such as video, music, entertainment, etc. With the rapid growth of information, content and data in multimedia information, users are facing more and more multimedia information. How to accurately show users their favorite content and screen out multimedia information that meets users’ needs is now a difficult problem.

At present, a method of studying the matching degree between users’ interests and attributes of multimedia information is usually adopted, and the predicted multimedia information with the highest matching degree is recommended to the user, thereby improving user experience. However, in the actual scenarios, users’ interests are not single and unchangeable. Users’ interests may evolve gradually in the long-term development, or users may present unique interests in different time periods and different scenarios in the short term. Therefore, the predicted multimedia information can’t meet the current interests of users, which further affects users’ experience.

(2) Key points of technical application

According to the multimedia information recommendation method, the related device and the computer storage medium described in this application, the user information of the target user, including the basic information of the target user, the historical sequence information of the multimedia information obtained by the target user and the multimedia information set, is obtained; then, the basic information of all target users, the historical sequence information of multimedia information obtained by the target users and the multimedia information set are input into the multimedia information recommendation model to obtain the click probability of each piece of multimedia information by the target users; among them, the multimedia information recommendation model is obtained by training the neural network model with the basic information of a plurality of training sample users, the historical sequence information of multimedia information acquired by the training sample users, the training sample multimedia information set and the real preference multimedia information of the training sample users; finally, a set of recommended multimedia information is generated according to the click probability of each piece of multimedia information by the target user. This application combines the historical sequence information of the multimedia information obtained by the user to determine the click probability of each piece of multimedia information by the user, so as to achieve the purpose of accurately recommending to the user the multimedia information more in line with the current interests of the user.

- 31 -

Chapter 3 Project Implementation Management

In the planning and management of project implementation, the R&D unit of this project is committed to building a systematic strategic management system. On the premise of the rapid progress of Internet technology and the improvement of competitive means, the external environment is becoming more and more complicated, the changes are accelerating, the uncertain factors are increasing, competition is fiercer, and the enterprise decision-making is becoming more and more complicated, all of which highlights the importance of strategic management. Therefore, enterprises need to build a systematic strategic management system to form systematic strategic management. The key and difficulty of building a strategic management system are to establish an effective internal and external environmental monitoring system. A team of professional strategic analysts will analyze the first-hand information obtained from the monitoring, make forward-looking and systematic analysis and planning on macro-environment, industry competition, market space, own resources and business capabilities, and work out clear and feasible short-term, medium-term and long-term operation objectives and career development directions of the enterprise.

- 32 -

3.1 Project management mode

The effectiveness of project management is directly related to the success or failure of the whole project, especially the research and development of new technology application projects related to the Internet, which is difficult both at home and abroad, and requires successful project management. This project management is carried out in line with the principles of the Project Management Institute and in combination with the characteristics of Internet projects on IT system development and creative design.

The management mode of this project mainly consists of three components:

3.1.1 Definition and organization of the project

(1) The overall demand of the project, customer background introduction and scheme composition;

(2) Definition of project scope;

(3) Composition structure, roles and responsibilities of the project team;

(4) Purpose and work target to be achieved by the project team;

(5) Internal coordination and independent management of the project team.

3.1.2 Plan of the Project

(1) Breaking down the work;

(2) Formulating a preliminary project implementation schedule;

(3) Balancing the project implementation schedule, project scope and resources;

(4) Risk management plan prediction and measures control;

(5) Cost control.

3.1.3 Tracking management of the project

(1) Collecting project status information;

(2) Analyzing the implementation schedule of the project, the scope of the project and the use of resources;

(3) Project progress report: generally once a week;

(4) Project document records: meeting logs, records and various memos;

(5) Project quality and customer satisfaction tracking;

(6) Summary after the completion of the project.

- 33 -

3.2 Project implementation method

The implementation method of a project is the premise and key element to ensure the successful completion of the project. It needs to effectively coordinate various professionals to participate in the management and allocation of resources in an organized and planned way, and ensure the completion of the project as required in terms of time and quality to the greatest extent.

This project is a continuous and intersecting implementation process of development and service application. The technical achievements of one phase are the basis of the next phase. They are interrelated and interact with each other, and organically constitute the implementation process of the whole project. Therefore, according to different tasks in different phases, this project dynamically allocates resources for implementation, and then combines with the professional knowledge of professionals, so that the project can be completed according to the corresponding processes. Processes and corresponding tasks of each phase of project implementation are:

(1) Planning and definition

The purpose of the planning and definition phase is to accurately grasp the business purpose of customers and establish the scope, integrity and operational implementation of the project. This includes a review of the customer’s business strategy; confirm, record and prioritize the list of requirements, and propose the draft of system architecture, select project members, integrate the project team and arrange the project plan according to the characteristics of the project.

(2) Analysis and design

After obtaining the project objectives, scope and high-level list of requirements and other results, a more detailed analysis and design is carried out in terms of functionality, system architecture technology and visual creativity, which are recorded one by one for discussion and improvement by both parties. If necessary, make a prototype or demonstration system to test the design concept. Then, complete the function development, interactive information and interface design in a targeted manner according to this design.

(3) Coding

Technical staff will code and develop the designed function results. Technical achievements will be put into use in phases.

- 34 -

(4) Test and acceptance

Test includes two parts: function test and performance test. Test results should be recorded in detail, and all technical and normative knowledge that customers must master and understand should be transferred during the test process to ensure that customers know how to operate and maintain the system. After the test, the Service Recipient will issue the acceptance certificate.

(5) Maintenance and management

In addition to the necessary monitoring and maintenance for the running system to ensure its normal operation, the more important task in the management and maintenance phase is to test the actual system performance from the actual operating system; find the parts of the system that need to be improved and upgraded during operation; and measure and compare the success of the system against business objectives and requirements. Organize all such information into a plan for future enhancement and upgrade of the system.

- 35 -