Content recommendation method and device

Abstract

The invention discloses a content recommendation method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the following steps: collecting real-time behavior data of a user, and generating real-time characteristics of the user based on the real-time behavior data; according to the real-time characteristics of the user, obtaining target recall content corresponding to the real-time characteristics of the user from a pre-configured recall database; and performing content recommendation on the user based on the target recall content. According to the embodiment, the real-time behavior data of the user can be collected, the corresponding user real-time features are generated, the content related to the recent behavior of the user is recalled through the user real-time features and actively pushed to the user, the real-time responsiveness is high, the content recommendation result is well matched with the latest requirement of the user, and therefore the user experience feeling is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, and in particular, to a content recommendation method and device. BACKGROUND

[0002] With the gradual rise of the Internet, Internet companies provide content information platforms to meet the needs of users to obtain information, and recommend interesting content to users. At present, the mainstream information platform is to analyze user behavior and interest preferences through the information flow recommendation mode, and to make personalized recommendations for users. However, the existing information flow recommendation mode is a pull mode, that is, the user needs to actively pull down to trigger the call of the recommendation service, so as to recommend real-time interesting content to the user. The above-mentioned mode can be triggered only by relying on the related operation of the user, and has poor real-time responsiveness and poor adaptability of the content recommendation result to the latest needs of the user, so that the user experience is poor. SUMMARY

[0003] Therefore, the embodiments of the present application provide a content recommendation method and device, which can collect real-time behavior data of a user, generate corresponding real-time features of the user, recall content related to the latest behavior of the user by using the real-time features of the user, and actively push the content to the user, so as to have strong real-time responsiveness, good adaptability of the content recommendation result to the latest needs of the user, and improved user experience.

[0004] To achieve the above-mentioned purpose, according to an aspect of an embodiment of the present application, a content recommendation method is provided, comprising:

[0005] Collecting real-time behavior data of a user, and generating real-time features of the user based on the real-time behavior data;

[0006] According to the real-time features of the user, obtaining target recall content corresponding to the real-time features of the user from a pre-configured recall database;

[0007] Performing content recommendation on the user based on the target recall content.

[0008] Optionally, according to the real-time features of the user, obtaining target recall content corresponding to the real-time features of the user from a pre-configured recall database, comprises:

[0009] Matching the real-time features of the user with content indexes in the recall database to obtain a candidate content identifier set corresponding to the real-time features of the user;

[0010] According to the candidate content identifier set, obtaining content detail data corresponding to the candidate content identifier set from a content cache, and generating candidate recall content;

[0011] Performing filtering processing on the candidate recall content, and sorting the candidate recall content after the filtering processing, and determining the target recall content according to a sorting result.

[0012] Optionally, the content recommendation is made to the user based on the target recall content, comprising:

[0013] The user recommendation content is constructed based on the target recall content.

[0014] The user recommendation content is stored in a user recommendation cache queue, and the content recommendation is made to the user based on the user recommendation cache queue.

[0015] Optionally, the content recommendation is made to the user based on the user recommendation cache queue, comprising:

[0016] In response to monitoring that the user is in a long connection subscription state, the user recommendation content is obtained from the user recommendation cache queue based on a preset push condition;

[0017] The user recommendation content is pushed to the terminal device of the user through a long connection channel.

[0018] The method further comprises:

[0019] In response to receiving a receiving success message fed back by the terminal device, a user recommendation content identifier corresponding to the receiving success message is determined.

[0020] The user recommendation content corresponding to the user recommendation content identifier is deleted from the user recommendation cache queue.

[0021] The pre-configured content exposure record sequence is updated based on the user recommendation content identifier, so that the content recommendation is made to the user according to the updated content exposure record sequence.

[0022] Optionally, after the content recommendation is made to the user based on the user recommendation cache queue, the method further comprises:

[0023] In response to receiving negative feedback information fed back by the user, the user recommendation content corresponding to the negative feedback information is determined.

[0024] The user negative feedback report is generated based on the user recommendation content corresponding to the negative feedback information.

[0025] The user recommendation cache queue is filtered according to the user negative feedback report, so that the content recommendation is made to the user based on the filtered user recommendation cache queue.

[0026] Optionally, the method further comprises:

[0027] The business content information corresponding to the business scenario is obtained.

[0028] The recall database is updated based on the business content information, so that the target recall content is obtained from the updated recall database.

[0029] According to a second aspect of an embodiment of the present application, a content recommendation device is provided, comprising:

[0030] The data acquisition module is used to collect real-time user behavior data and generate real-time user characteristics based on the real-time behavior data.

[0031] The acquisition module is used to retrieve target recall content corresponding to the user's real-time characteristics from a pre-configured recall database based on the user's real-time characteristics.

[0032] The recommendation module is used to recommend content to users based on the targeted content recall.

[0033] According to a third aspect of the present invention, an electronic device is provided, comprising:

[0034] One or more processors;

[0035] Memory, used to store one or more programs.

[0036] When one or more programs are executed by one or more processors, the one or more processors implement the methods of any of the above embodiments.

[0037] According to a fourth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method of any of the above embodiments.

[0038] According to a fifth aspect of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the method of any of the above embodiments.

[0039] One embodiment of the above invention has the following advantages or beneficial effects: by collecting users' real-time behavior data, real-time user features are generated based on the real-time behavior data; according to the users' real-time features, target recall content corresponding to the users' real-time features is obtained from a pre-configured recall database; and content is recommended to users based on the target recall content. This embodiment can collect users' real-time behavior data and generate corresponding real-time user features, use the users' real-time features to recall content related to the users' recent behavior, and actively push it to the users. It has strong real-time responsiveness, and the content recommendation results are well adapted to the users' latest needs, thereby improving the user experience.

[0040] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0041] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0042] Figure 1 This is a schematic diagram of the main flow of the content recommendation method according to an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of the main flow of a content recommendation method according to a preferred embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the main modules of a content recommendation device according to an embodiment of the present invention;

[0045] Figure 4 This is an exemplary system architecture diagram in which embodiments of the present invention can be applied;

[0046] Figure 5 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation

[0047] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0048] It should be noted that the acquisition, storage, and application of personal information involved in the embodiments of the present invention comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0049] With the rise of the internet, internet companies provide content and information platforms to meet users' information needs and recommend content of interest. Currently, mainstream information platforms analyze user behavior and interests through news feed recommendations to provide personalized recommendations. However, the existing news feed recommendation model is a pull-based model, requiring users to actively pull down to trigger the recommendation service before real-time content of interest is presented. This method relies on user actions, resulting in poor real-time responsiveness and difficulty in adapting content recommendations to users' latest needs, leading to a poor user experience.

[0050] In view of this, according to one aspect of the present invention, a content recommendation method is provided.

[0051] Figure 1 This is a schematic diagram of the main flow of the content recommendation method according to an embodiment of the present invention. Figure 1 As shown, the content recommendation method according to an embodiment of the present invention includes the following steps S101 to S103.

[0052] Step S101: Collect real-time user behavior data and generate real-time user characteristics based on the real-time behavior data.

[0053] This embodiment continuously monitors and collects real-time interactive information generated by users in various business scenarios during application use. Specifically, real-time behavioral data includes dynamic interactive data such as click behavior, browsing behavior, page dwell behavior, like behavior, comment behavior, and sharing behavior on business pages such as lending scenarios, consumption scenarios, wealth management scenarios, check-in scenarios, content detail scenarios, fund channels, and insurance channels. The above data can reflect the user's real-time focus and changes in interest within the application. User real-time features are structured feature information used to characterize the user's real-time interests. For example, user real-time features include feature fields that can be used to characterize changes in user interests, such as event identifiers, behavior types, topic tags, content tags, content identifiers, followed author information, and behavior matching rules.

[0054] Specifically, real-time user behavior data can be collected based on a client-side behavior reporting mechanism. When the client detects a user's click, browsing, or interaction, it immediately encapsulates the behavior data into a behavior event and reports it to the server via event tracking. The behavior event is then written to the behavior message queue in real time without affecting the user experience. Alternatively, data can be collected based on business system event streams. When a user triggers a business action (such as entering a channel, viewing content details, or interacting with content), each business system writes the corresponding behavior event to a unified business event stream, subscribes to this stream, and processes the behavior event into a unified format of behavior message data, thereby achieving behavior data aggregation across business scenarios.

[0055] After collecting real-time user behavior data, real-time user features can be generated using Flink tasks. Flink tasks receive real-time user behavior events such as clicks, browsing, and interactions from a behavior message queue across various business scenarios. These events are then parsed, cleaned, and have their fields completed. Real-time data features, such as topic tags and content tags, are extracted based on pre-defined behavior matching rules. Alternatively, real-time behavior data can be input into a pre-trained feature extraction model to obtain real-time user features.

[0056] Step S102: Based on the user's real-time characteristics, retrieve the target recall content corresponding to the user's real-time characteristics from the pre-configured recall database.

[0057] The recall database stores a set of recommendable content resources, including inverted and forward index databases. The inverted index database is indexed according to different dimensions such as tags, topics, authors, and content types, enabling rapid filtering of content identifiers that meet matching criteria after receiving real-time user characteristics. The forward index database stores content details corresponding to content identifiers, including structured information such as content title, content tags, content type, click-through rate, like rate, and content publication time. Targeted recall content is a set of content that matches a user's current behavior, interests, or content association rules. For example, a user clicking on fund-related content will generate fund-related topic tags, and a user following an author will generate author-related author features, and so on.

[0058] Retrieving target content from the recall database can be done using an index query approach. When real-time user characteristics (such as topic tags, content tags, or followed author characteristics) are received, the index fields are directly matched, and the corresponding content identifier set is retrieved from the inverted index database. Then, based on the content identifier set, the corresponding content details are retrieved from the forward index database to obtain the target content. Alternatively, a rule-based matching and real-time processing approach can be used. The corresponding recall source can be determined based on the trigger type in the user's real-time characteristics. For example, "user clicks on content events" can be mapped to popular content recall sources, and "users follow authors events" can be mapped to author-related content recall sources. Based on the recall sources, content that meets timeliness, quality, or weighting conditions is filtered in the recall database to form the target content.

[0059] Step S103: Recommend content to users based on the target recall content.

[0060] In this embodiment, a pre-trained ranking model can be used to sort the target recalled content, thereby determining the target recommended content that better matches the current user's interest intensity and decision-making stage, and then presenting the target recommended content to the user. Alternatively, based on rule-based policy configuration, the target recalled content can be filtered according to business rules, including but not limited to factors such as content quality, content exposure control, content cold start degree, and recommendation priority set by the business side. Based on the filtering results, a final recommendation list is generated, where each recommended item is arranged in the list according to its business weight and user matching degree, thereby achieving content recommendation to the user.

[0061] This embodiment collects real-time user behavior data and generates real-time user features based on this data. Based on these features, it retrieves target content corresponding to the user's real-time features from a pre-configured recall database and recommends content to the user based on this target content. This embodiment can collect real-time user behavior data, generate corresponding real-time user features, recall content related to the user's recent behavior using these features, and proactively push this content to the user. It exhibits strong real-time responsiveness, and the content recommendation results are well-matched to the user's latest needs, thereby improving the user experience.

[0062] Optionally, based on the user's real-time characteristics, the target recall content corresponding to the user's real-time characteristics is obtained from a pre-configured recall database, including: matching the user's real-time characteristics with the content index in the recall database to obtain a set of candidate content identifiers corresponding to the user's real-time characteristics; based on the set of candidate content identifiers, obtaining the content details data corresponding to the set of candidate content identifiers from the content cache, and generating candidate recall content; filtering the candidate recall content, sorting the filtered candidate recall content, and determining the target recall content based on the sorting result.

[0063] This embodiment first uses real-time user characteristics to perform feature-level matching with a pre-built content index in the recall database. Keyword indexes, feature vector indexes, or behavioral pattern indexes in the index structure are used to quickly locate the range of content related to the current user's real-time characteristics, thus obtaining a set of candidate content identifiers. Then, based on the candidate content identifier set, corresponding content details are extracted from the content cache. The content cache stores structured data, feature vectors, and real-time updated content interaction metrics. Candidate recall content is efficiently generated by querying the candidate content identifier set one by one. After obtaining the candidate recall content, a filtering process is further performed. This filtering process can eliminate candidate recall content based on business rules, content quality metrics, content validity, exposure control strategies, and user-sensitive category restrictions to ensure that content that meets recommendation requirements is obtained. The filtered candidate recall content enters the sorting process. A sorting model or strategy comprehensively considers real-time user characteristics, content characteristics, and the degree of matching between content and user, etc., to prioritize the candidate recall content, and the final target recall content is determined based on the sorting results.

[0064] In addition to the above implementation methods, a recall method based on a real-time event triggering mechanism can also be adopted. When user behavior triggers specific rules, a preset set of content is directly retrieved from the recall database according to the rules. This approach enables efficient recall based on real-time user characteristics, allowing recommended content to quickly respond to changes in user behavior, thereby improving recommendation relevance and user experience.

[0065] Optionally, recommending content to users based on the target recall content includes: constructing user-recommended content based on the target recall content; storing the user-recommended content in a user-recommended cache queue; and recommending content to users based on the user-recommended cache queue.

[0066] In this embodiment, firstly, content details, content features, display styles, delivery strategy parameters, and their association information with real-time user characteristics are extracted from the target recall content. Then, the above content is formatted according to the recommendation strategy to generate user-recommended content suitable for the current user scenario. After the user-recommended content is formed, it is written into a user recommendation cache queue. The user recommendation cache queue is used to temporarily store recommendation content to be delivered to users, organizing the recommendation content according to time order or priority strategy to ensure that the recommendation results can be consumed in real time. Subsequently, content recommendations are output to users based on the user recommendation cache queue. This can be achieved by pushing the user recommendation content in the user recommendation cache queue to users through push channels, page rendering interfaces, or instant response mechanisms, realizing a real-time content recommendation service for users. In addition, when constructing user-recommended content, dynamic content can be generated directly by calling a multimodal generation model based on the target recall content, thereby improving the richness of the recommended content. Alternatively, a queue-free direct push method can be used to output user-recommended content immediately after generation to further reduce recommendation response latency; or the user recommendation cache queue can be bound to user behavior events, allowing the recommended content to be dynamically distributed according to the order of behavior triggering.

[0067] This embodiment can improve recommendation efficiency while ensuring the accuracy of recommended content. By introducing a user recommendation cache queue, the recommendation output becomes more stable and faster. It can also dynamically adjust the recommended content based on real-time user behavior, thereby improving the real-time performance and user experience of the recommendation system.

[0068] Optionally, content recommendation for users based on a user recommendation cache queue includes: in response to detecting that a user is in a long-connection subscription state, retrieving user-recommended content from the user recommendation cache queue based on preset push conditions; and pushing the user-recommended content to the user's terminal device through a long-connection channel. The method further includes: in response to receiving a successful reception message from the terminal device, determining the user-recommended content identifier corresponding to the successful reception message; deleting the user-recommended content corresponding to the user-recommended content identifier from the user recommendation cache queue; and updating a pre-configured content exposure record sequence based on the user-recommended content identifier, so as to recommend content to the user according to the updated content exposure record sequence.

[0069] In this embodiment, the system continuously monitors whether the user is in a long-connection subscription state. When a long-connection subscription state is detected, user-recommended content suitable for the current push timing is retrieved from the user recommendation cache queue according to preset push conditions. These push conditions can be determined based on time intervals, user interaction behavior, content exposure rhythm, or content priority strategies. After the push conditions are met, the retrieved user-recommended content is pushed to the user's terminal device through the long-connection channel, enabling the terminal device to receive new content in real time without the user actively refreshing the page, thereby realizing a proactive push-based content recommendation service.

[0070] In addition, the system monitors whether the terminal device sends a successful reception message. Upon receiving such a message, it identifies the corresponding user-recommended content identifier and removes the user-recommended content corresponding to that identifier from the user-recommended content cache queue to maintain the real-time performance and freshness of the cache queue. Based on this identifier, the system updates the pre-configured content exposure record sequence. By recording already exposed content, duplicate pushes can be avoided in subsequent recommendation processes. Furthermore, dynamic adjustments can be made based on content exposure frequency, content type, and content relevance to the user, thereby improving the rationality of subsequent recommendation decisions. Besides the above implementation, an event-driven push method can be used, retrieving recommended content from the user-recommended content cache queue and pushing it when the user triggers a specific behavioral event (such as page dwell or scrolling to the bottom). Alternatively, short polling or lightweight push channels can be used as alternatives to long connections to adapt to terminal devices with poor network environments or those that do not support long connections.

[0071] This embodiment enables real-time content push with low latency, allowing users to receive content highly relevant to their interests without actively refreshing the page. At the same time, by maintaining the content exposure record sequence, it improves the accuracy of the recommendation logic and the diversity of content distribution, thereby enhancing the user experience and the effectiveness of content recommendation.

[0072] Optionally, after recommending content to users based on the user recommendation cache queue, the method further includes: in response to receiving negative feedback information from users, determining the user recommendation content corresponding to the negative feedback information; generating a user negative feedback report based on the user recommendation content corresponding to the negative feedback information; filtering the user recommendation cache queue according to the user negative feedback report, and recommending content to users based on the filtered user recommendation cache queue.

[0073] After recommending content to users based on the user recommendation cache queue, the system continuously monitors feedback from users' terminal devices. When negative feedback is detected, the system first determines the corresponding recommended content based on the content identifier or context information contained in the negative feedback. This includes, but is not limited to, specific recommended content that the user explicitly dislikes, closed the pop-up, clicked the "dislike" button, or expressed rejection in other ways. After identifying the corresponding recommended content, a user negative feedback report is constructed based on this content. This report may include information such as the type of content the user is not interested in, content feature tags, user interaction records, and the time the negative feedback occurred, comprehensively depicting the user's reasons for rejecting this type of content or changes in preference. Subsequently, the user recommendation cache queue is filtered based on the user negative feedback report, removing recommended content of the same type, tag, or similar characteristics as the negative feedback content to ensure that the content retained in the cache queue better matches user preferences. For the filtered user recommendation cache queue, subsequent push notifications continue to recommend content to users based on the filtered cache queue, thereby dynamically adjusting the recommendation results to ensure that the recommended content always maintains real-time relevance to user interests. In addition, a recall phase can be triggered after negative feedback occurs to retrieve new candidate content, thereby proactively replenishing the user recommendation cache queue and maintaining the diversity of content supply.

[0074] This embodiment enables real-time response to negative user feedback, quickly identifies content that users are not interested in, and adjusts subsequent recommendation results in a timely manner. This improves the consistency between recommended content and user interests, reduces invalid pushes, enhances user experience, and strengthens the sensitivity and adaptability of the recommendation system to changes in user intent, thereby improving content distribution efficiency and overall recommendation effectiveness.

[0075] Optionally, the method further includes: obtaining business content information corresponding to the business scenario; updating the recall database based on the business content information, so as to obtain the target recall content from the updated recall database.

[0076] In the recommendation process, the recall database can be dynamically updated based on actual business conditions to ensure that the recall results continuously reflect the latest business content. In this implementation, business content information corresponding to the business scenario is first obtained. This information can originate from business content generation services, content review services, data synchronization tasks, or content management platforms, and includes attributes such as unique content identifiers, content topic tags, text information, interaction metrics data, and publication time. After obtaining the business content information, it is parsed and structured according to the needs of the business scenario, and the processed information is updated in the recall database. Update operations may include writing new content entries to the recall database, updating the tag attributes or statistical metrics of existing content entries, or removing expired content, etc., to ensure that the content stored in the recall database is always a valid and usable set of content for the current business scenario. Through the updated recall database, target recall content can be retrieved from the latest content set during recall, keeping the recommendation results sensitive to changes in business content in real time. In other implementations, the recall database can be periodically updated through scheduled tasks, or updates can be pushed in real time when business content changes based on trigger-based update logic.

[0077] This embodiment can effectively improve the freshness and coverage of the recall database, enabling timely capture of changes in business content, hot topic trends, and content quality during the recall phase. This further enhances the relevance and real-time nature of the target recall content, providing higher-quality candidate content for subsequent ranking and push, and improving the overall accuracy and user experience of the recommendation system.

[0078] Figure 2 This is a schematic diagram of the main flow of a content recommendation method according to a preferred embodiment of the present invention. Figure 2As shown, this embodiment proactively pushes content based on users' real-time behavior, improving the timeliness of content recommendations and the user browsing experience. First, real-time user behavior data is collected across various business scenarios within the financial app, including click behavior, interaction behavior, and content browsing behavior in scenarios such as lending, consumption, wealth management, and check-in. This real-time behavior data is processed by Flink tasks to generate real-time user features, which are then sent to a real-time user feature message queue for subsequent recommendation processing. Real-time user features include user identifiers, event identifiers, event matching rules, topic tags, content tags, content identifiers, and information about followed authors. The real-time user feature message queue is monitored, and corresponding content recall pools or databases are pre-generated based on different user behaviors. For example, a content recall pool is generated for authors followed by the user, or relevant content is recalled based on user interest tags. Content is stored using an Elasticsearch and Redis index architecture, where Elasticsearch is used for fast content retrieval via inverted index, and Redis cache stores detailed content information. Based on real-time user characteristics, candidate recall content is obtained from the recall database or content recall pool. After filtering and sorting, target recall content is generated, and the recommendation ranking model is called to calculate the fine ranking score and timeliness score to obtain high-quality user recommendation content.

[0079] In addition, before pushing user-recommended content to users, a security interface is called to perform content compliance verification to determine whether the user-recommended content is compliant. If the content does not comply with the rules, it is recorded in the non-compliance log for easy troubleshooting. Content that complies with the rules is stored in the user recommendation cache queue or the user's mailbox. The cache queue is designed as a fixed-length queue with a default capacity of three items to ensure the timeliness of the latest recommended content and its relevance to recent user behavior. When a user is in the information stream subscription state, user-recommended content is retrieved from the user recommendation cache queue based on preset push conditions and actively pushed to the user's terminal device through a long connection channel. After receiving the user-recommended content, the terminal device displays it in a pop-up window at the bottom and reports relevant user behavior tracking information for subsequent recommendation effect analysis and user behavior analysis. The successful reception message from the client triggers the server to delete the pushed content and update the content exposure record sequence to avoid duplicate pushes.

[0080] During the user feedback process, the terminal device can send negative feedback messages, such as clicking the "X" in a pop-up window. A user negative feedback report is generated based on this message, and the user recommendation cache queue is filtered based on this report to optimize subsequent recommendation content, improve recommendation relevance, and enhance user experience. Furthermore, dynamic updates to the recall database are supported. By obtaining business content information corresponding to the business scenario, newly added or updated content is synchronized to the recall database to ensure real-time consistency between the recalled content and the business content.

[0081] This embodiment constructs a complete real-time user behavior monitoring and content recommendation mechanism, realizing the ability to push content instantly after discovering new user interests, ensuring that the recommended content is highly relevant to the user's latest behavior, increasing user browsing time and interaction depth, while reducing the user's reliance on actively refreshing the information stream, and optimizing the real-time performance and accuracy of information stream recommendations.

[0082] According to a second aspect of the present invention, a content recommendation device is provided.

[0083] Figure 3 This is a schematic diagram of the main modules of a content recommendation device according to an embodiment of the present invention. Figure 3 As shown, the content recommendation device 300 includes:

[0084] The data acquisition module 301 is used to collect real-time behavioral data of users and generate real-time user characteristics based on the real-time behavioral data.

[0085] The acquisition module 302 is used to retrieve target recall content corresponding to the user's real-time characteristics from a pre-configured recall database based on the user's real-time characteristics.

[0086] Recommendation module 303 is used to recommend content to users based on the target recalled content.

[0087] Optionally, the acquisition module 302 is also used for:

[0088] The user's real-time features are matched with the content index in the recall database to obtain a set of candidate content identifiers corresponding to the user's real-time features;

[0089] Based on the candidate content identifier set, retrieve the content detail data corresponding to the candidate content identifier set from the content cache, and generate candidate recall content;

[0090] The candidate recall content is filtered and sorted. The target recall content is determined based on the sorting results.

[0091] Optionally, the recommended module 303 is also used for:

[0092] Build user recommendation content based on targeted recall content;

[0093] User-recommended content is stored in a user recommendation cache queue, and content recommendations are made to users based on the user recommendation cache queue.

[0094] Optionally, the recommended module 303 is also used for:

[0095] In response to the detection that a user is in a long-connection subscription state, the system retrieves user-recommended content from the user recommendation cache queue based on preset push conditions.

[0096] Push user-recommended content to the user's terminal device via a long connection channel;

[0097] The content recommendation device 300 also includes a deletion module, which is used for:

[0098] In response to receiving a successful reception message from the terminal device, determine the user-recommended content identifier corresponding to the successful reception message;

[0099] Remove the user-recommended content corresponding to the user-recommended content identifier from the user-recommended content cache queue;

[0100] The pre-configured content exposure record sequence is updated based on the user's recommended content identifier, so as to recommend content to the user according to the updated content exposure record sequence.

[0101] Optionally, the content recommendation device 300 also includes a filtering module, which is used for:

[0102] In response to receiving negative feedback from users, determine the user recommendation content corresponding to the negative feedback.

[0103] Generate a user negative feedback report based on the user recommendation content corresponding to the negative feedback information;

[0104] The user recommendation cache queue is filtered based on user negative feedback reports, and content recommendations are made to users based on the filtered user recommendation cache queue.

[0105] Optionally, the content recommendation device 300 also includes an update module, which is used for:

[0106] Obtain business content information corresponding to the business scenario;

[0107] The recall database is updated based on business content information in order to retrieve the target recall content from the updated recall database.

[0108] It should be noted that the specific implementation details of the content recommendation device of this invention have been described in detail in the content recommendation method above, so the details will not be repeated here.

[0109] According to a third aspect of the present invention, an electronic device is provided, comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in the first aspect of the present invention.

[0110] According to a fourth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method provided in the first aspect of the present invention.

[0111] According to a fifth aspect of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the method provided in the first aspect of the present invention.

[0112] Figure 4 An exemplary system architecture 400 is shown that can be applied to the content recommendation method or content recommendation apparatus of the present invention.

[0113] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, and 403, a network 404, and a server 405. Network 404 serves as the medium for providing communication links between terminal devices 401, 402, and 403 and server 405. Network 404 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0114] Users can use terminal devices 401, 402, and 403 to interact with server 405 via network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, and 403, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0115] Terminal devices 401, 402, and 403 can be various electronic devices with displays that support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0116] Server 405 can be a server that provides various services, such as a backend management server that supports shopping websites browsed by users using terminal devices 401, 402, and 403 (for example only). The backend management server can analyze and process data such as received content recommendation requests, and feed back the processing results (such as user-recommended content - for example only) to the terminal device.

[0117] It should be noted that the content recommendation method provided in this embodiment of the invention is generally run by server 405, and correspondingly, the content recommendation device is generally set in server 405.

[0118] It should be understood that Figure 4 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0119] The following is for reference. Figure 5 It shows a schematic diagram of the structure of a computer system 500 suitable for implementing a terminal device of the present invention. Figure 5The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0120] like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 502 or programs loaded from storage section 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the system 500. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0121] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. Drive 510 is also connected to I / O interface 505 as needed. Removable media 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0122] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is run by the central processing unit (CPU) 501, it performs the functions defined above in the system of this invention.

[0123] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0124] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more operable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually operate substantially in parallel, and they may sometimes operate in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0125] The modules described in the embodiments of the present invention can be implemented in software or hardware. The described modules can also be housed in a processor; for example, a processor may be described as including a collection module, an acquisition module, and a recommendation module. The names of these modules do not necessarily limit the module itself; for example, a recommendation module may also be described as "a module for recommending content to users based on targeted content recall."

[0126] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs that, when executed by the device, cause the device to include: collecting real-time behavioral data of a user; generating real-time user features based on the real-time behavioral data; retrieving target recall content corresponding to the user's real-time features from a pre-configured recall database based on the user's real-time features; and recommending content to the user based on the target recall content.

[0127] The computer program product provided in this embodiment of the invention includes a computer program that, when executed by a processor, implements the content recommendation method in this embodiment of the invention.

[0128] The technical solution of this invention has the following advantages or beneficial effects: by collecting real-time user behavior data, real-time user features are generated based on the real-time behavior data; target recall content corresponding to the user's real-time features is obtained from a pre-configured recall database according to the user's real-time features; and content is recommended to the user based on the target recall content. This embodiment can collect real-time user behavior data and generate corresponding real-time user features, recall content related to the user's recent behavior using the user's real-time features, and actively push it to the user. It has strong real-time responsiveness, and the content recommendation results are well adapted to the user's latest needs, thereby improving the user experience.

[0129] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0130] It should be noted that the acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

Claims

1. A content recommendation method, characterized in that, include: Collect real-time user behavior data and generate real-time user characteristics based on the real-time behavior data; Based on the user's real-time characteristics, retrieve the target recall content corresponding to the user's real-time characteristics from the pre-configured recall database; Content recommendations are made to the user based on the target recall content.

2. The method according to claim 1, characterized in that, Based on the user's real-time characteristics, retrieve target recall content corresponding to the user's real-time characteristics from a pre-configured recall database, including: The user's real-time features are matched with the content index in the recall database to obtain a set of candidate content identifiers corresponding to the user's real-time features; Based on the candidate content identifier set, retrieve the content detail data corresponding to the candidate content identifier set from the content cache, and generate candidate recall content; The candidate recall content is filtered, and the filtered candidate recall content is sorted. The target recall content is determined based on the sorting result.

3. The method according to claim 1, characterized in that, Based on the target recall content, content recommendations are made to the user, including: User recommendation content is constructed based on the target recall content; The user-recommended content is stored in a user-recommended cache queue, and content recommendations are made to the user based on the user-recommended cache queue.

4. The method according to claim 3, characterized in that, Based on the user recommendation cache queue, content recommendation is performed on the user, including: In response to detecting that a user is in a long-connection subscription state, the system retrieves user-recommended content from the user recommendation cache queue based on preset push conditions; The user-recommended content is pushed to the user's terminal device via a long connection channel; The method further includes: In response to receiving a success message from the terminal device, the user recommended content identifier corresponding to the success message is determined; Delete the user recommendation content corresponding to the user recommendation content identifier from the user recommendation cache queue; The pre-configured content exposure record sequence is updated based on the user-recommended content identifier, so as to recommend content to the user according to the updated content exposure record sequence.

5. The method according to claim 3, characterized in that, After recommending content to the user based on the user recommendation cache queue, the method further includes: In response to receiving negative feedback from a user, determine the user-recommended content corresponding to the negative feedback; A user negative feedback report is generated based on the user recommendation content corresponding to the negative feedback information. The user recommendation cache queue is filtered based on the user negative feedback report, and content recommendations are made to the user based on the filtered user recommendation cache queue.

6. The method according to claim 1, characterized in that, The method further includes: Obtain business content information corresponding to the business scenario; The recall database is updated based on the business content information to obtain the target recall content from the updated recall database.

7. A content recommendation device, characterized in that, include: The data acquisition module is used to collect real-time behavioral data of users and generate real-time user characteristics based on the real-time behavioral data. The acquisition module is used to acquire target recall content corresponding to the user's real-time characteristics from a pre-configured recall database based on the user's real-time characteristics. The recommendation module is used to recommend content to the user based on the target recalled content.

8. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images