Recommendation information display method and device, electronic equipment, medium and program product

Abstract

The invention provides a recommendation information display method and device, electronic equipment, a medium and a program product, and relates to the technical field of computers.The method comprises the steps that a client displays a search result for a search keyword on a first page, and when target input used for quitting display of the search result is received, the search result is displayed on the first page; sending a search request to a server; the server side responds to the search request of the client side, generates target recommendation information according to a user portrait of a target user, and sends the target recommendation information to the client side; the client displays a second page, wherein the second page comprises the target recommendation information; wherein the second page is a page triggered and displayed by the client based on the target input of the target user on the first page or the search input of the target user on the home page of the client. The method can be used for improving the click rate and the conversion rate of the content application platform.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and particularly to a method, apparatus, electronic device, medium and program product for displaying recommended information. Background Art

[0002] With the continuous growth of users' demand for consuming long video content, online streaming platforms have increasingly emphasized the optimization of search functions. In order to increase the click-through rate and conversion rate of various types of information in an application, related technologies usually display recommended information by setting a hot search module on the search interface, so as to attract viewing users to trigger the recommended information. However, in the current search link, after a viewing user enters a keyword in the search interface of the client to query relevant information, the client jumps from the search interface to the query result display interface. When the viewing user selects to return or re-click the search box, the client usually retains the original search keyword or directly reuses the keyword, directly inheriting the context information of the search result page. In this process, the user's interest and behavior data are not effectively captured and utilized, resulting in a decrease in potential playback opportunities and user stickiness, and limiting the platform's ability to increase the click-through rate and conversion rate. Summary of the Invention

[0003] The present disclosure provides a method, apparatus, electronic device, medium and program product for displaying recommended information, which is used to solve the problem that in related technologies, the user's interest and behavior data are not effectively captured and utilized, resulting in a decrease in potential playback opportunities and user stickiness, and the platform's click-through rate and conversion rate are restricted.

[0004] In a first aspect, the present application provides a method for displaying recommended information, which is applied to a client. The method includes:

[0005] The client displays search results for a search keyword on a first page, and in response to receiving a target input for exiting the display of the search results, sends a search request to a server.

[0006] The server responds to the search request of the client, generates target recommended information according to the user profile of the target user, and sends the target recommended information to the client.

[0007] The client displays a second page, and the second page includes the target recommended information.

[0008] Wherein, the second page is a page triggered and displayed by the client based on the target input of the target user on the first page or the search input on the home page of the client.

[0009] In a second aspect, the present application further provides a system for displaying recommended information, including:

[0010] A client is used to display search results for search keywords on a first page. When receiving a target input for exiting the display of the search results, it sends a search request to a server and displays a second page, where the second page includes the target recommendation information.

[0011] The server is used to respond to the search request of the client, generate target recommendation information according to the user profile of the target user, and send the target recommendation information to the client.

[0012] Wherein, the second page is a page triggered and displayed by the client based on the target input of the target user on the first page or the search input on the home page of the client.

[0013] In a third aspect, the present application provides an electronic device, including a processor, a memory, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the method described in the first aspect.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the method described in the first aspect.

[0015] In a fifth aspect, the present application provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the steps of the method described in the first aspect.

[0016] In the present application, the client initiates a search request when exiting the first page for displaying search results. The server then generates target recommendation information in combination with the user profile. The second page of the client is dynamically triggered by the target input or the home page search, and the target recommendation information determined based on the user profile of the target user is displayed on the second page. In this way, in the process of the target user returning from the first page interface for displaying search results to the second page, by effectively utilizing the returned second page, the present application embodiment realizes the dynamic recommendation of platform content and the accurate information matching, thereby enhancing the user's attention and click-through rate on the platform content and improving the conversion rate of the content. Description of the Drawings

[0017] Figure 1 is a flowchart of a method for displaying recommendation information provided by an embodiment of the present application;

[0018] Figure 2 is a schematic diagram of the first page provided by an embodiment of the present application;

[0019] Figure 3 is one of the schematic diagrams of the second page provided by an embodiment of the present application;

[0020] Figure 4 It is the second schematic diagram of the second page provided in the embodiments of the present application;

[0021] Figure 5 It is the third schematic diagram of the second page provided in the embodiments of the present application;

[0022] Figure 6 It is a schematic structural diagram of a recommended information display system provided in the embodiments of the present application;

[0023] Figure 7 It is a schematic structural diagram of an electronic device provided in the embodiments of the present application. Detailed implementation manners

[0024] Next, the technical solutions in the embodiments of the present disclosure will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some, but not all, of the embodiments of the present disclosure. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present disclosure without creative efforts shall fall within the protection scope of the present disclosure.

[0025] The embodiments of the present disclosure provide a method for displaying recommended information, as Figure 1 shown, the method for displaying recommended information includes:

[0026] Step 11: The client displays search results for a search keyword on a first page, and sends a search request to the server when receiving a target input for exiting the display of the search results.

[0027] Among them, the first page is any page in the client for displaying search results. For example, the first page is the search result page of a video playback client, the search result page of a music playback client, the search result page of a shopping client, the search result page of an e-book client, etc.

[0028] The search results can be determined according to the search keyword input by the target user into the client, and the search results can be information related to the search keyword. The content displayed on the first page corresponding to different search keywords is different.

[0029] Exemplarily, in a certain video application platform, as Figure 2 shown, if the search keyword is "Journey to the West", then the first page, as the display interface of the search results, displays content such as TV dramas, movies, animations, etc. related to Journey to the West, including all related works of different versions and different work types.

[0030] In addition, search keywords can be input into the client through various methods such as voice, gestures, button touch control, and text input. The search keywords can be understood as the keywords input by the target user into the client on their own initiative, or they can be multiple keywords pre-set by the client on the search page, and the user can select one of the multiple keywords as the search keyword, that is, the clicked keyword.

[0031] In this application, the target user can obtain search results through the search operation on the search control in the search interface of the client, or enter the search interface of the client through voice input, and then obtain search results by voice inputting search keywords. At this time, the client jumps from the search interface to the first page, and the first page presents search results associated with the search keywords.

[0032] It should be noted that the target input received by the client can be understood as an input instruction obtained by the client through conversion based on the operations of the target user to exit the first page or the search result display page on the first page.

[0033] In one case, when the target user clicks the exit or return control in the first page, the client converts the click operation on the exit control into an operation instruction to exit the first page or the search result display page, thereby obtaining the target input.

[0034] In another case, when the target user clicks the search control in the first page, the client converts the click operation on the search control into an operation instruction to jump to the search main interface, thereby obtaining the target input. Among them, the search control can be a control for inputting search keywords, and it is a user interface (User Interface, UI) element through which the user operates to input search keywords or trigger related behaviors. For example, the search control can be a search box, a search button, a voice search input button, a touch gesture area, etc., and this application does not make specific limitations.

[0035] Step 12: The server responds to the search request of the client, generates target recommendation information according to the user portrait of the target user, and sends the target recommendation information to the client.

[0036] In the application, the user profile of the target user can include at least three feature dimensions, namely the basic information, behavioral characteristics, and interest tags of the target user. Among them, when the user permits and it is compliant, the basic information can include the age, gender, geographical location, etc. of the target user. The behavioral characteristics can be data generated based on the historical operations, viewings, usages, etc. of the target user on the client side. For example, descriptive characteristics such as user preference types (such as ancient costume dramas, wuxia films, etc.), "preference for short videos", etc. The interest tags can be formed by using methods such as Natural Language Processing (NLP) to refine the themes and keywords of the historical viewed videos, so as to achieve content matching through the interest tags.

[0037] In this application, the user profiles pre-stored on the server can include the basic information, behavioral characteristics, historical viewing records, interest tags, etc. of the users, enabling the server to more accurately match the preferences of the users with the help of the historical user profiles.

[0038] Among them, the server will comprehensively analyze the user profile and the search request based on algorithms such as deep learning, collaborative filtering, and regression models to identify the content that the user may be interested in. For example, if a user searches for "science fiction movies" and their user profile shows a preference for this type of content, the server will focus on the popular movies and recently released related works in this category to determine the target recommendation information.

[0039] In some embodiments, the client processes the user's behavioral data, such as clicks, viewings, likes, etc., and quickly feeds it back to the server side, and the server side accordingly updates the user profile. Specifically, when the target user and the client complete a new interaction, the client sends the behavioral data of the target user, and the server analyzes the behavioral data of the target user through algorithms and responds to the changes in habits and interests without waiting for regular batch updates. For the content information that the target user may be interested in, the server can cover the features and attributes of all content through the established recommendation matrix and match the user profile with these contents through calculation. At the same time, the server can also use machine learning models to obtain trends from users who have viewed long videos, so as to infer the potential interest of users in similar long video programs.

[0040] In this application, when a user conducts a search on the client side, search requests generated by either entering keywords, voice commands, or other means will be pushed to the server. By receiving these search requests, the server can understand the needs and interest orientations of the currently active users, providing data support for subsequent recommendations. The server receives the search requests sent by the client, and the search requests may include user identifiers for locating user portraits, search keywords, context information (such as the type of exit action (e.g., return key, swipe to exit), timestamp, device type, etc.), and trigger sources (used to distinguish whether it is an exit operation on the first page or a search input on the home page). Subsequently, the data carried in the search requests is processed, the search keywords are cleaned (such as removing stop words) and semantically analyzed (such as word segmentation, entity recognition), or user behavior sequences (such as click records before and after the search) can also be extracted to capture changes in user interests.

[0041] In a specific embodiment, the server can retrieve the user portrait from a database or cache through the user identifier, and can input the user portrait, search keywords, context features, etc. into a sorting model to obtain the target recommendation information output by the sorting model. Among them, the sorting model can adopt a content-based matching model, such as using TF-IDF, Word2Vec, etc. to calculate the semantic similarity between the search keywords and the product / media content, or can also adopt a collaborative filtering model to recommend the media content they interacted with by finding user groups similar to the current target user. Alternatively, a user behavior feedback model can also be adopted to preferentially recall content related to the user's recent clicks / searches. In addition, popular supplements can also be set, that is, in the case of cold start or low-active users, popular or promotional content is inserted.

[0042] Step 13, the client displays a second page, and the second page includes the target recommendation information; wherein, the second page is a page displayed by the client based on the target input of the target user on the first page or a search input on the home page of the client.

[0043] The above-mentioned second page can be understood as a search intermediate interface, that is, the second page is the interface entered by the user after exiting from the first page (search result page) (such as Figure 3As shown, it can carry multiple functions and information, such as a search recommendation module (a module that provides search keywords for target users), a hot search list module (a module for displaying popular recommended content), and a good movie search module (i.e., the module in this application for displaying target recommended information) to enhance the user experience, improve the content consumption conversion rate, and further explore user interests. The second page can provide a search box for users, allowing them to re-enter new keywords for searching, and also supports continued searching through various methods such as voice or gestures, providing convenience for users with different habits. The second page can also display popular recommended content, such as "similar works", "newly launched", "high user ratings", etc., to help users quickly judge content relevance.

[0044] In this application, when exiting the first page of the client for displaying search results, target recommended information determined based on the user profile of the target user is displayed on the second page of the client, achieving a rapid transition from search results to personalized recommendations. Since the displayed target recommended information is determined based on the updated user profile, it can efficiently grasp user interests, thereby providing more accurate and attractive content, increasing user attention to platform content, improving the content consumption conversion rate of the client, increasing the probability of users clicking on recommended content, promoting users' viewing decisions, and ultimately enhancing the traffic of the client platform.

[0045] In a specific embodiment, the target user is currently browsing a video playback client. The target user enters the target keyword "Four Great Classical Novels" on the second page, and the client receives the target input. The page switches from the second page to the first page, and at this time, the first page will display multiple film and television works related to "Four Great Classical Novels", such as works like *Dream of the Red Chamber*, *Water Margin*, *Romance of the Three Kingdoms*, etc. However, the target user does not jump to the viewing interface through the work links displayed on the first page, or the user hopes to exit the first page after watching, or the target user hopes to redefine the search keyword to obtain more accurate search results.

[0046] At this time, the target user can exit the first page by clicking the "Back" button or the search control. The client detects that the target user has clicked the "Back" button, that is, it receives the target input, and the client page jumps from the first page to the second page. The target recommended information displayed on the second page is generated by the server based on the user profile of the target user. Exemplarily, if the user profile shows that the target user prefers works adapted from Chinese opera and classic literature, the target recommended information may include video links of recently popular TV series, movies, or original stage plays related to works such as *Dream of the Red Chamber* and *Romance of the Three Kingdoms*, as Figure 4 shown. Thus, the generated target recommended information is based on the matching of aesthetic elements and content features to increase the user's viewing interest.

[0047] It can be understood that in response to a trigger instruction for an information search control on the third page, the client displays media hot search information in the content display area within the second page. Among them, the third page can be the channel main page entered when the client is launched, such as the user interface for multimedia channel recommendations in a video client, or the main interface for entering various application ports in a shopping client. The second page can display both target recommendation information (as shown in Figure 3 ), and media hot search information (as shown in Figure 5 ). However, when entering the second page from the third page, the content display area of the second page displays media hot search information. When a target input for exiting the display of search results is taken from the first page, the second page will display the target recommendation interface.

[0048] Thus, the embodiment of the present application realizes the full utilization of the link of returning to the search page after exiting the search result page, enhances the user's attention to the platform content, improves the consumption conversion rate of the client content, and increases the probability of the user clicking on the recommended content.

[0049] In one embodiment, step 12 includes:

[0050] Step 121, the server responds to the search request of the client and obtains the behavior data of the target user; wherein, the behavior data is carried in the search request, and the behavior data includes the search keyword, page click behavior, and page stay duration;

[0051] Step 122, the server determines the user portrait of the target user based on the behavior data;

[0052] Step 123, the server inputs the behavior data and the user portrait of the target user into a preset model to obtain the target content information predicted by the preset model that the target user has an interaction tendency;

[0053] Step 124, the server generates the target recommendation information according to the user portrait of the target user, the target content information, and the search keyword.

[0054] It can be understood that the user portrait of the target user can be pre-constructed and stored by the server, and the data sources for constructing the user portrait can include the target user's historical viewing records, user preferences, viewing duration, etc.

[0055] In the application, the client can record the behavior of the target user before exiting the search results (such as page stay duration, click location, sliding speed, etc.) through data logging, and then report it to the server log system through the SDK. The server extracts the semantic features of the search keywords. For example, "Dream of the Red Chamber" can be recognized as "TV drama / historical drama / sentimental drama", and combines the context to supplement the intention. For example, if the target user browsed the search result page before exiting but did not click on any media content within the search result page, it may imply that the user did not find satisfactory content.

[0056] Among them, the historical viewing records can include all the videos watched by the user, including the content information of each video, such as video title, type, duration, rating, etc., which can help identify the user's long-term preferences. The user preferences can be obtained directly or indirectly based on forms such as questionnaires, user feedback, selected favorites, or liked videos. The viewing duration records the viewing duration of the user on each video, which is used to judge the user's preference for certain types or specific content.

[0057] The search keyword can be the keyword entered by the target user in the search box on the first page or the second page to understand the user's current needs and interests. The search keyword helps to identify the preference trend of the target user.

[0058] The click situation can be understood as the frequency of the links, buttons, and other interface elements clicked by the target user, which can include the displayed content (such as video title, thumbnail) and the displayed location to infer the type of content that the user is more interested in.

[0059] The stay time on the content interface can be understood as the access duration of the target user on different pages. The stay time of the target user on the content interface helps to evaluate the attractiveness of the content on the page and infer the reasons for the loss of the target user during the browsing process.

[0060] As for the page return operation, it can be understood as the operation of the target user returning from one page to the previous page, which can indicate that the target user did not find relevant content or was not satisfied with the content displayed on the current page. Therefore, the page return operation can be used as a reference for improving the user experience.

[0061] In this application, the prediction models that the server can adopt include regression models, time series analysis, collaborative filtering, deep learning models, etc. Personalized recommendations are achieved through multi-dimensional data fusion and dynamic weight adjustment. Among them, regression models can be used to predict the weights of each label in the user profile. For example, a logistic regression model can be used to predict the probability that a user clicks on a certain category. Time series analysis can be used to capture the periodic changes in interests. For example, the Prophet model analyzes the periodicity of users' drama-watching behaviors. Collaborative filtering models can solve the cold start problem and achieve similarity recommendations based on the preferences of users of the same type. Deep learning models can achieve complex interest representation learning. For example, the user interest evolution is modeled through the Deep Interest Network (DIN).

[0062] In a specific embodiment, the server calculates the cosine similarity between the user interest label and the content label based on the input user profile, target content information, and search keywords (such as the matching degree between the user "digital enthusiast" and the product "active noise-canceling headphones"). It can also adjust the weight of entertainment content according to the current time (such as weekend evenings), and knowledge graphs can also be used to expand keywords.

[0063] The above user profile may include the static attributes of the target user (such as age, region, etc.), dynamic interest labels (such as ancient costume enthusiast: 0.8), and user behavior sequences, such as the search and click records in the last 2 minutes. The above target content information may include the metadata of the candidate content (title, category, label), and context associations (such as the relevance between "Journey to the West" and "Water Margin"). The above search keywords may be semantic vectors processed by NLP (such as BERT embeddings), and intent classifications (such as emotion, action, children, etc.).

[0064] In the application, potential relevant candidate content can be quickly screened through a multi-way parallel strategy. The specific methods include: recall based on the user profile, that is, matching the user label and the content label (weighted by TF-IDF). For example, high-scoring ancient costume dramas are recalled through the label of the user being an ancient costume enthusiast. Recall based on the search term, such as using Elasticsearch for semantic search and expanding synonyms. Collaborative filtering recall can find similar dramas according to the user's historical click content through Item-CF, etc., or the latest dramas of the previously watched director can be recalled. Behavioral recall, such as capturing the click behaviors in the last 5 minutes through Flink stream processing and preferentially recalling media content of the same category. Popular / trend recall can also be adopted, counting the content with a soaring click-through rate recently (weighted by time decay) and displaying newly released dramas, etc.

[0065] Furthermore, a machine learning model can be used to accurately rank the recalled content above. The input features on the model side can include user-side features such as long-term interest vectors, user behavior sequences, user attributes, etc., content-side features such as category Embedding, text semantic vectors, statistical features (such as CTR, conversion rate), etc., and context features such as device type (mobile / web), network environment, geographical location, etc. Exemplarily, a 2-fold weight can be applied to the content related to the search keyword, or negative feedback punishment can be used, that is, if the user has ignored the same type of content, a fixed value is deducted from the score. Diversity control can also be performed to ensure a balanced category distribution on the same page (such as at most 3 ancient costume dramas) through a greedy algorithm.

[0066] In addition, this application can intervene in the sorting results with business rules, adopt a deduplication strategy, such as filtering the content that has been exposed within 24 hours, or adopt commercial insertion, that is, insert advertisements at fixed positions, or adopt cold start protection. If the score of new content enters the top 3, it is forcibly promoted to the top 10, or adopt sensitive content filtering, such as blocking restricted-level content according to the user's age.

[0067] In this way, through these models, it is convenient for the server to predict the content preferences of the target user based on historical user behavior data, behavior data, and the current search keyword, etc., and recommend works of related types and similar content according to this content preference, realizing the dynamic recommendation of platform content and accurate information matching.

[0068] In one embodiment, step 122 includes:

[0069] Step 1221, the server determines the user portrait of the target user based on the behavior data, including:

[0070] Step 1222, the server extracts the feature points corresponding to multiple media contents in the behavior data, and the feature points are used to characterize the interaction tendency of the target user with respect to the multiple media contents;

[0071] Step 1223, the server updates the historical user portrait of the target user stored in advance according to the feature points to obtain the updated user portrait of the target user.

[0072] In this application, the target recommendation information displayed on the second page can be predicted by the server. The server constructs a user portrait through the historical behavior data of the target user uploaded by the client, and updates the user portrait according to the behavior data of the target user, the search keyword input by the target user, etc.

[0073] In a specific embodiment, sequence modeling can be used to take the target user's most recent N behaviors (such as searching for "ancient costume drama" → "ancient costume fantasy" → "Journey to the West") as sequence inputs to predict the next point of interest. Exemplarily, the attention mechanism of Transformer can be used to capture the dependencies between behaviors. For example, "campus" may be associated with "campus love".

[0074] In another specific embodiment, the portrait tag library is updated based on the behavior type. For example, when the user searches for "Journey to the West" and stays for 3 minutes, the corresponding weight is added to the "ancient costume fantasy" tag. Or, according to the user's rapid swiping operation when exiting, the weights of the tags related to the content on the current page can be reduced. In addition, collaborative filtering can be adopted for supplementation. If the user's behavior is sparse, the user is embedded into the vector space through the User2Vec model, and the behaviors of similar users are used to supplement the tags. For example, if user A is similar to user B, some of user B's tags can be partially assigned to user A.

[0075] In yet another specific embodiment, models such as BERT can be used to map search terms to the interest dimension. For example, "Dream of the Red Chamber" → the Embedding vector is close to "ancient costume drama / family emotions". It is also possible to expand associated interests in combination with a knowledge graph. For example, "Ne Zha" can be expanded to "animation". In addition, if the current search keyword is too different from the historical portrait, for example, a target user with a mother and baby tag suddenly searches for "e-sports mouse", an anomaly detection model is triggered to determine whether it is a temporary interest or account sharing.

[0076] Thus, through the above embodiments, the behavior data of the target user can be analyzed to predict the current interest point of the target user.

[0077] Exemplarily, the behavior data of the target user obtained by the server includes {user ID: 123, behavior: search for "Journey to the West" → exit, time: 2025-3-1 14:2:00}, and the historical portrait of this user is {tag: ancient costume lover (0.8)}. The relevance between "Journey to the West" and the historical tag is calculated through the term frequency–inverse document frequency (TF-IDF) (such as the relevance of "ancient costume lover" is 0.9). The Long Short-Term Memory (LSTM) model is used to analyze the recent behavior sequence (3 searches in the past 1 hour are all related to ancient costume dramas), and it is predicted that the interest is concentrated in ancient costume type dramas. Subsequently, a dynamic interest vector can be generated through the DIN model, and the similarity with the candidate content Embedding is calculated. Finally, the portrait tag weights are updated (the "ancient costume lover" changes from 0.8 to 0.85, and a new tag "fantasy drama lover" is added with a weight of 0.7).

[0078] In the embodiments of the present application, by dynamically extracting interaction feature points in user behavior and updating the user profile, it is possible to accurately capture interest changes, improve the degree of recommendation personalization, and combine historical data and behavior data to enhance the adaptability of the model to the evolution of user preferences, effectively improve recommendation relevance, reduce information overload, and improve user experience and content consumption efficiency.

[0079] In one embodiment, step 12 includes:

[0080] Step 125, the server responds to the search request and obtains context data of the search result, where the context data includes at least one of the following: the item features not triggered in the search result, the semantic parsing result of the search keyword;

[0081] Step 126, the server generates target recommendation information based on the user profile of the target user and the context data through a preset negative feedback recommendation strategy; wherein, the negative feedback recommendation strategy is used to exclude content types corresponding to the context data.

[0082] It can be understood that the extraction of the above untriggered item features can be the interaction behavior of the user on the first page (search result page), such as an exposed but unclicked item, that is, the content that the user slides past but does not stay / click, or a comparison behavior, that is, the user quickly returns after clicking item A and continues to browse (implies dissatisfaction with A). In this regard, common features of untriggered items are extracted (such as the number of episodes of a drama > 36, the director = "X", the category = "action").

[0083] The above semantic parsing of the search keyword can include intention classification, for example, using the BERT model to determine whether the search term belongs to "TV drama", "movie" or "variety show", and entity disambiguation, for example, parsing "Ne Zha" as "animation" or "cartoon".

[0084] In a specific embodiment, dynamic filtering rules are constructed according to the untriggered item features. For example, if the user skips all items with the number of episodes of a drama greater than 36, the generated rule is range NOT IN("36-40","40+"). If the user ignores movie media content, content_type = movie can be filtered. A hybrid recommendation strategy can be set. Intervention can be set in the recall stage to block content corresponding to blacklist features, or the semantic parsing result can be used to limit the candidate set range. For example, under the intention of "TV drama", movies, variety shows, etc. are excluded. Specifically, the sorting model optimization can be achieved by adding negative feedback signals to the CTR model features, and the similarity with blacklist features can also be reduced through the Maximal Marginal Relevance (MMR) algorithm, and finally target recommendation information is generated.

[0085] In another embodiment of the present application, the search request includes a search keyword and target indication information, where the target indication information can indicate whether to update the user profile of the target user. The target indication information can be preset by the target user. For example, the target user sets that the target indication information indicates to update or not to update the user profile in different time periods.

[0086] In one case, when a search instruction input by the target user is received on the first page, for example, when the client receives a search instruction such as a search keyword, a search voice instruction, or a search gesture of the target user, the client can generate a search request before the first page is displayed and send the search request to the server.

[0087] In another case, when a search instruction input by the target user is received on the first page, the client displays the first page related to the search instruction. After the first page is displayed, the client generates a search request and sends the search request to the server. In this way, it can be ensured that before exiting the first page, the client can obtain the target recommendation information predicted by the server for the generated search request, and when an instruction to exit the first page is received, the target recommendation information can be promptly and quickly displayed on the second page.

[0088] For example, within a relatively long time period, such as the first month of each week, it is allowed that the user profile is more sensitive to new input responses. If the target user frequently searches for new content categories during this period, even if there is only a slight deviation, the user profile can be updated. For example, when the target user shows stable preferences, the client may lower the update threshold and only update when the input keywords are significantly different from the past. Or, during a specific node or event (such as a promotion season), the target user may try new product categories, so the function of updating the user profile is enabled by default. At this time, the target indication information indicates to update the user profile, encouraging attempts and explorations.

[0089] In some embodiments, the meaning represented by the target indication information can also be determined according to the similarity between the search keyword input by the target user and the historical input keywords.

[0090] Exemplarily, when a user searches for "sports shoes" on a certain shopping client, if the system notices that multiple sports brand products emerge in the user's purchase history, the target indication information may be set to "no", and there is no need to update the user profile. However, if the target user further searches for "luxury bags", such an input is quite different from the purchase pattern, and the target indication information should be set to "yes" to re-evaluate the target user's consumption behavior and preferences, thereby updating the user profile.

[0091] In an application, the client can obtain the user's behavior data from multiple channels (websites, mobile applications, social media, etc.) and centrally store it in a big data platform. For example, the client uses data pipelines (such as Apache Kafka, Apache Flink) to collect and transmit data to the server. Subsequently, the server uses clustering algorithms (such as K-means, hierarchical clustering, etc.) to divide the target users into corresponding groups in order to identify the same type of behavior patterns.

[0092] In one embodiment, step 11 includes:

[0093] Step 111, when the client receives the target input, generate the search request according to the search keyword and the running mode of the client;

[0094] Step 112, wherein the running mode of the client includes a first mode and a second mode. The first mode supports updating the user profile of the target user, and the second mode does not support updating the user profile of the target user.

[0095] In this application, the generation of the search request is closely related to the running mode of the client. The search instruction input by the target user and the running mode of the client can determine how the client processes the search request and whether to update the user profile.

[0096] Among them, the target user can input a specific search keyword on the first page to trigger relevant content searches. For example, the user may input "popular TV series" or "new movies". Subsequently, the client can construct a complete search request based on the search instruction input by the target user and the current running mode of the client. This request will include the search keyword input by the user and additional running mode information.

[0097] The above first mode can be understood as that in this mode, if the target user conducts a search, the client can continuously update its user profile according to the input of the target user. This means that new search behaviors will be incorporated into the user's history, thereby adjusting the recommended content predicted by the server. For example, in this mode, after the user inputs "romance movies", if this type has not been included in the user profile, the user profile will be updated to include such content.

[0098] The second mode can be understood as that in this mode, even if the target user makes a new search input, the client will not change the user profile. This mode is applicable to scenarios where the user hopes to maintain personal preferences unchanged or has privacy concerns, ensuring that the previously collected user behavior data and other information will not be changed due to changes in the user's search behavior.

[0099] Exemplarily, the target user opens the video client and asks about the currently popular TV dramas. In the first mode, such as the adult mode, the client allows the user's portrait to reflect the new interest in these dramas. If the user has not watched them before but then searches for related dramas multiple times, the client will update its data accordingly. In the second mode, such as the teenage mode, even if there are frequent searches, the client still maintains the previous data and does not update it to ensure stability.

[0100] For another example, when the user browses the e-commerce client and enters "running shoes", in the first mode, such as the master mode, if the user decides to purchase, the user portrait is updated to add this record so that future recommendations can include more similar products. In the second mode, such as the guest mode, even if the same search is entered, the client still uses the old user portrait and only recommends the previously counted products without considering the new search behavior.

[0101] It can be seen that the embodiments of the present application effectively meet the different needs of different users for personalization and stability, and at the same time improve user engagement and trust by combining the generation of search requests with the operating mode of the client, providing different operating modes for users to choose the experience that best suits their needs, or by the client identifying different application scenarios and using the corresponding operating modes. By allowing users to select the operating mode according to their needs, the client can effectively adapt to various user behaviors and improve the user experience and satisfaction.

[0102] In one embodiment, the method further includes:

[0103] Step 01: Display a target control on the first page of the client;

[0104] Step 02: When the client receives a status adjustment instruction from the target user for the target control, adjust the current first state of the target control to a second state, or adjust the current second state of the target control to the first state;

[0105] Wherein, the first state represents updating the user portrait of the target user, and the second state represents not updating the user portrait of the target user.

[0106] The above target control is used to allow the target user to directly control their search behavior and the impact on the user portrait. Through the status adjustment instruction for the target control, the target user can independently choose whether they want to update their user portrait to better match the current interests and needs of the target user.

[0107] Among them, the target control can be a UI component displayed on the first page, such as components like switches, buttons, or checkboxes. The target user uses the target control to indicate to the client how to process data related to their search behavior.

[0108] In this application, when the user operates on the target control (such as clicking, swiping, or toggling a switch), this operation can be converted by the client into a "status adjustment instruction". The client will change the state of the target control according to the received status adjustment instruction and accordingly decide whether to update the user profile.

[0109] In some embodiments, the first state is used to represent updating the user profile. It can be understood that the target user hopes that the client can update their user profile in a timely manner according to the search keywords and specific application behaviors they input, so as to obtain more personalized content recommendations. For example, the user may find that their interests have changed, so they actively choose to update. The second state is used to represent not updating the user profile. It can be understood that the target user hopes to maintain the stability of the current user profile, which usually occurs when the user clearly knows that their interests will not change or for privacy reasons. At this time, even if the target user conducts a new search, the target user will not attempt to update the user profile.

[0110] Exemplarily, in an e-commerce client, the target user is searching for summer clothes. The target user can click the "Update Personalized Recommendations" button (i.e., the target control) to choose whether they hope the client updates their shopping preferences in the user profile according to the current search. If they choose not to update, they will continue to be dominated by their favorite products without being disturbed by new search situations.

[0111] In another example, the target user is browsing a certain video client and switches the target control from the closed state to the "on" state, which means the target user hopes that the client can update their viewing records in a timely manner and make adjustments to the new viewing habits. If it remains "closed", no matter how the user searches for movies, their user profile will not change.

[0112] It can be seen that the embodiments of this application are beneficial to enhancing user trust and satisfaction and improving the user's satisfaction with the personalized recommendation effect by giving the user control over personal data association. When the user chooses to update the profile, the client can more sensitively capture the user's interest changes, so as to provide more accurate recommended content. When the user does not want to update the profile, it ensures the efficiency of the current recommendation, reduces the frequent fluctuations of the user profile caused by accidental factors, and avoids interference from noisy data. Thus, by establishing the interaction mechanism in the embodiments of this application, the user's sense of participation is enhanced, and the user's stickiness to the platform is improved.

[0113] In the embodiments of the present application, an efficient chain from user behavior to personalized recommendation is realized, enabling the online platform to better serve its users and enhancing the user experience and platform benefits. Among them, by using the user's search intent and the stored user portrait information for content recommendation, the relevance of the recommended content is improved, thereby promoting the user's decision-making process. In addition, each search request of the target user can reflect the latest user needs or interest changes, enabling the server to continuously and efficiently adapt to the user's preferences, shortening the time for the user to find the ideal content, and enhancing the content conversion rate and click-through rate of the client side.

[0114] In one embodiment, the search request includes the search keyword and target indication information, and the target indication information is used to indicate whether to update the user portrait of the target user;

[0115] Predicting the target recommendation information based on the search request and the pre-stored user portrait of the target user includes:

[0116] In the case where the target indication information indicates to update the user portrait of the target user, updating the user portrait of the target user based on the search keyword;

[0117] Predicting the target recommendation information based on the updated user portrait of the target user.

[0118] It can be understood that all the implementation manners in the above embodiments can refer to the relevant descriptions in the foregoing recommendation information display method applied to the client side. To avoid repeated description, this embodiment will not be elaborated herein.

[0119] See Figure 6 , the embodiments of the present disclosure provide a recommendation information display system, as Figure 6 shown, the recommendation information display system 20 includes:

[0120] The client 21 is configured to display search results for the search keyword on the first page, and send a search request to the server side when receiving a target input for exiting the display of the search results, and display a second page, where the second page includes the target recommendation information;

[0121] The server side 22 is configured to respond to the search request of the client, generate target recommendation information according to the user portrait of the target user, and send the target recommendation information to the client;

[0122] Among them, the second page is a page triggered and displayed by the client based on the target input of the target user on the first page or the search input on the home page of the client.

[0123] In one embodiment, the server side 22 is specifically configured to:

[0124] In response to the search request of the client, obtain the behavior data of the target user; wherein, the behavior data is carried in the search request, and the behavior data includes the search keyword, page click behavior, and page stay duration.

[0125] Based on the behavior data, determine the user profile of the target user.

[0126] Input the behavior data and the user profile of the target user into a preset model to obtain the target content information with an interaction tendency predicted by the preset model for the target user.

[0127] Generate the target recommendation information according to the user profile of the target user, the target content information, and the search keyword.

[0128] In one embodiment, the server 22 is specifically configured to:

[0129] Extract the feature points corresponding to multiple media contents in the behavior data, where the feature points are used to characterize the interaction tendency of the target user for the multiple media contents.

[0130] According to the feature points, update the historical user profile of the target user stored in advance to obtain the updated user profile of the target user.

[0131] In one embodiment, the server is configured to:

[0132] In response to the search request, obtain the context data of the search result, where the context data includes at least one of the following: the item features not triggered in the search result, the semantic analysis result of the search keyword.

[0133] The server generates the target recommendation information based on the user profile of the target user and the context data through a preset negative feedback recommendation strategy; wherein, the negative feedback recommendation strategy is used to exclude the content types corresponding to the context data.

[0134] In one embodiment, the client is configured to:

[0135] When receiving the target input, generate the search request according to the search keyword and the running mode of the client.

[0136] Wherein, the running mode of the client includes a first mode and a second mode. The first mode supports updating the user profile of the target user, and the second mode does not support updating the user profile of the target user.

[0137] In one embodiment, the system is further configured to:

[0138] Display a target control on a first page of the client;

[0139] When the client receives a status adjustment instruction for the target control from the target user, adjust the current first state of the target control to a second state, or adjust the current second state of the target control to the first state;

[0140] Wherein, the first state represents updating the user profile of the target user, and the second state represents not updating the user profile of the target user.

[0141] The recommendation information display system 20 provided by the embodiments of the present disclosure can implement each process in the above-mentioned recommendation information display method embodiments as shown in Figure 1 For the sake of avoiding repetition, it will not be elaborated here.

[0142] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device and a readable storage medium.

[0143] Figure 7 FIG. shows a schematic block diagram of an exemplary electronic device 30 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0144] As Figure 7 shown, the device 30 includes a computing unit 31, which can execute various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 32 or a computer program loaded from a storage unit 38 into a random access memory (RAM) 33. In the RAM 33, various programs and data required for the operation of the device 30 can also be stored. The computing unit 31, the ROM 32, and the RAM 33 are connected to each other through a bus 34. An input / output (I / O) interface 35 is also connected to the bus 34.

[0145] Multiple components in device 30 are connected to I / O interface 35, including: input unit 36, such as a keyboard, mouse, etc.; output unit 37, such as various types of displays, speakers, etc.; storage unit 38, such as a disk, optical disc, etc.; and communication unit 39, such as a network card, modem, wireless communication transceiver, etc. Communication unit 39 allows device 30 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0146] Computing unit 31 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of computing unit 31 include but are not limited to a central processing unit (CPU), a graphic process unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Computing unit 31 executes the various methods and processes described above, such as the recommended information display method. For example, in some embodiments, the recommended information display method can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 38. In some embodiments, part or all of the computer program can be loaded and / or installed onto device 30 via ROM 32 and / or communication unit 39. When the computer program is loaded into RAM 33 and executed by computing unit 31, one or more steps of the recommended information display method described above can be executed. Alternatively, in other embodiments, computing unit 31 can be configured to execute the recommended information display method in any other suitable way (e.g., by means of firmware).

[0147] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field-programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on chip (SOC) systems, complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which may be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0148] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing device, such that the program codes, when executed by the processor or controller, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine, or entirely on the remote machine or server.

[0149] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0150] As used herein, the term "machine-readable medium" refers to any computer program product, apparatus, and / or device (e.g., a magnetic disk, an optical disk, a memory, a programmable logic device (PLD)) that provides machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal that provides machine instructions and / or data to a programmable processor.

[0151] To provide for interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic, speech, or tactile input).

[0152] The systems and techniques described herein can be implemented in a computing system that includes a back-end component (e.g., as a data server), or a computing system that includes a middleware component (e.g., an application server), or a computing system that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or in a computing system that includes any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0153] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of the client and the server is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0154] The embodiments of the present application also provide a computer program product, including computer instructions, which when executed by a processor, implement the above Figure 1 or Figure 6Each process of the method embodiment shown can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0155] It should be understood that various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution disclosed in this disclosure can be achieved. There is no limitation herein.

[0156] The above specific embodiments do not constitute a limitation on the protection scope of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub - combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the protection scope of this disclosure.

Claims

1. A method for displaying recommended information, characterized in that: include: The client displays search results for the search keyword on the first page, and sends a search request to the server when receiving a target input for exiting the display of the search results; The server responds to the search request of the client, generates target recommendation information according to the user portrait of the target user, and sends the target recommendation information to the client; The client displays a second page, wherein the second page includes the target recommendation information; The second page is a page that is triggered and displayed by the client based on the target input of the target user on the first page or the search input on the homepage of the client.

2. The method according to claim 1, characterized in that The server generates target recommendation information according to the user profile of the target user in response to the search request of the client, and sends the target recommendation information to the client, including: The server responds to the search request of the client and obtains the behavior data of the target user; wherein the search request carries the behavior data, and the behavior data includes the search keyword, page click behavior and page dwell time; The server determines a user profile of the target user based on the behavior data; The server inputs the behavior data and the user portrait of the target user into a preset model to obtain target content information that the preset model predicts the target user has an interactive tendency; The server generates the target recommendation information according to the user portrait of the target user, the target content information and the search keywords.

3. The method according to claim 2, characterized in that The server determines the user profile of the target user based on the behavior data, including: The server extracts feature points corresponding to the multiple media contents in the behavior data, where the feature points are used to characterize the target user's interaction tendency with respect to the multiple media contents; The server updates the pre-stored historical user portrait of the target user according to the feature points to obtain an updated user portrait of the target user.

4. The method according to claim 1, characterized in that: The server generates target recommendation information according to the user profile of the target user in response to the search request of the client, including: The server obtains, in response to the search request, context data of the search result, wherein the context data includes at least one of the following: untriggered item features in the search result, and semantic analysis results of the search keyword; The server generates target recommendation information based on the user portrait of the target user and the context data through a preset negative feedback recommendation strategy; wherein the negative feedback recommendation strategy is used to exclude content types corresponding to the context data.

5. The method according to any one of claims 1 to 4, characterized in that: When the client receives a target input for exiting display of the search results, the client sends a search request to the server, including: When receiving the target input, the client generates the search request according to the search keyword and the operation mode of the client; The operation mode of the client includes a first mode and a second mode, the first mode supports updating the user portrait of the target user, and the second mode does not support updating the user portrait of the target user.

6. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Displaying a target control on a first page of the client; When receiving the state adjustment instruction of the target control from the target user, the client adjusts the current first state of the target control to the second state, or adjusts the current second state of the target control to the first state; The first state represents updating the user profile of the target user, and the second state represents not updating the user profile of the target user.

7. A recommendation information display system, characterized in that: include: The client is configured to display search results for a search keyword on a first page, and upon receiving a target input for exiting the display of the search results, send a search request to a server and display a second page, wherein the second page includes the target recommendation information; The server is used to respond to the search request of the client, generate target recommendation information according to the user portrait of the target user, and send the target recommendation information to the client; The second page is a page that is triggered and displayed by the client based on the target input of the target user on the first page or the search input on the homepage of the client.

8. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the method for displaying recommendation information as claimed in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the recommendation information display method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the method for displaying recommendation information as claimed in any one of claims 1 to 6.

Images