Recommendation method and device, electronic equipment and storage medium

Abstract

The invention provides a recommendation method and device, electronic equipment and a storage medium, and relates to the technical field of computers, in particular to the technical field of artificial intelligence such as large models and deep learning. According to the specific implementation scheme, using information and a user portrait of a target user in a current reference time period are input into a large model, and a plurality of interest tags of the target user output by the large model and the confidence coefficient of each interest tag are obtained; according to the correlation degree between each candidate content in the candidate content library and the interest tag, obtaining reference content from the candidate content library; determining a plurality of first to-be-recommended contents from all the reference contents based on the correlation degree corresponding to each reference content and the confidence coefficient of the associated interest tag; determining second to-be-recommended content based on a current recommendation target of the system; and sending the second to-be-recommended content and the at least one first to-be-recommended content to the client of the target user.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, especially to artificial intelligence technologies such as large models and deep learning, and specifically relates to a recommendation method, apparatus, electronic device, and storage medium. Background Art

[0002] Light users of a recommendation system refer to users who have visited the recommendation system for a relatively small number of days and have also viewed relatively few texts, images, videos, etc., such as new users. Since light users are relatively easy to lose, reducing the churn rate of light users is of great significance for increasing the user base of the recommendation system, improving the commercial conversion efficiency, and enhancing competitiveness. The key means to reduce the churn rate of light users is to achieve more accurate content recommendations for light users, so as to improve the satisfaction of users with the recommendation system and increase the retention efficiency. Summary of the Invention

[0003] The present disclosure aims to at least solve one of the technical problems in the related art to some extent.

[0004] To this end, the purpose of the present disclosure is to propose a recommendation method, apparatus, electronic device, and storage medium. By forcibly recommending the recommended content determined based on the user's interest points, the accuracy of the recommended content is improved when recommending content to light users, providing conditions for improving the conversion rate of light users.

[0005] According to a first aspect of the present disclosure, a recommendation method is provided, including:

[0006] Input the usage information and user profile of a target user within the current reference period into a large model, and obtain multiple interest tags of the target user output by the large model and the confidence level of each interest tag;

[0007] Obtain reference content from the candidate content library according to the relevance between each candidate content in the candidate content library and the interest tags;

[0008] Determine multiple first content to be recommended from all the reference content based on the relevance corresponding to each reference content and the confidence level of the associated interest tags;

[0009] Determine second content to be recommended based on the current recommendation target of the system;

[0010] Send the second content to be recommended and at least one of the first content to be recommended to the client of the target user.

[0011] According to a second aspect of the present disclosure, a recommendation apparatus is provided, including:

[0012] A processing module, configured to input the usage information and user profile of a target user within the current reference period into a large model, and obtain multiple interest tags of the target user output by the large model and the confidence level of each interest tag;

[0013] An acquisition module, configured to obtain reference content from the candidate content library according to the association degree between each candidate content in the candidate content library and the interest tags;

[0014] A first determination module, configured to determine multiple first content to be recommended from all the reference content based on the association degree corresponding to each reference content and the confidence level of the associated interest tags;

[0015] A second determination module, configured to determine second content to be recommended based on the current recommendation target of the system;

[0016] A sending module, configured to send the second content to be recommended and at least one of the first content to be recommended to the client of the target user.

[0017] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0018] At least one processor; and

[0019] A memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the recommendation method as described in the first aspect.

[0021] According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, and the computer instructions are used to cause a computer to execute the recommendation method as described in the first aspect.

[0022] According to a fifth aspect of the present disclosure, there is provided a computer program product, including computer instructions, and the computer instructions, when executed by a processor, implement the steps of the recommendation method as described in the first aspect.

[0023] The recommendation method, device, electronic device and storage medium provided by the present disclosure have the following beneficial effects:

[0024] By obtaining the content to be recommended related to the interest and the content to be recommended that meets the system recommendation goal according to the interest tags and confidence levels obtained from the user usage information and portrait analysis, and sending the content to be recommended determined by both methods to the client for display at the same time, where the content to be recommended sent contains at least one piece of content to be recommended related to the interest, it is possible to ensure that each piece of content recommended to the user must contain content related to the user's interest while achieving the system recommendation goal, improving the reliability and accuracy of content recommendation, and being beneficial to improving the satisfaction and retention rate of users, especially light users.

[0025] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the drawings. The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure, where:

[0027] Figure 1 is a flowchart of a recommendation method according to an embodiment of the present disclosure;

[0028] Figure 2 is a flowchart of a recommendation method according to another embodiment of the present disclosure;

[0029] Figure 3 is a flowchart of a recommendation method according to another embodiment of the present disclosure;

[0030] Figure 4 is a flowchart of a recommendation method according to another embodiment of the present disclosure;

[0031] Figure 5 is a flowchart of a recommendation method according to another embodiment of the present disclosure;

[0032] Figure 6 is a structural diagram of a recommendation device according to an embodiment of the present disclosure;

[0033] Figure 7 shows a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, descriptions of well-known functions and structures are omitted in the following description for clarity and conciseness.

[0035] The embodiments of the present disclosure relate to the field of artificial intelligence technologies such as large models and deep learning.

[0036] A large model is a type of artificial intelligence model with a large number of parameters constructed by an artificial neural network. Usually, it is first pre-trained on a large amount of data through self-supervised learning or semi-supervised learning, and then its performance and capabilities are further optimized through methods such as instruction tuning and human alignment. Large models have characteristics such as a large number of parameters, large training data, and large computing resources, and have the capabilities to solve general tasks, follow human instructions, and perform complex reasoning. The main categories of artificial intelligence large models include: large language models (LLMs), vision large models, multi-modal large models, and basic science large models, etc.

[0037] Deep learning is to learn the internal laws and representation levels of sample data, and the information obtained during these learning processes is very helpful for the interpretation of data such as text, images, and sounds. The ultimate goal of deep learning is to enable machines to have the ability to analyze and learn like humans, and be able to recognize data such as text, images, and sounds.

[0038] Artificial Intelligence, abbreviated as AI in English. It is a new technical science that studies, develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence.

[0039] In the technical solutions of the present disclosure, the processing of the collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0040] The following describes the recommendation method, device, electronic device, and storage medium of the embodiments of the present disclosure with reference to the accompanying drawings.

[0041] It should be noted that the execution subject of the recommendation method in this embodiment is a recommendation device, which can be implemented in software and / or hardware, and this device can be configured in an electronic device, and the electronic device can include but is not limited to terminals, server sides, etc.

[0042] Figure 1It is a schematic flowchart of a recommendation method proposed according to an embodiment of the present disclosure.

[0043] As Figure 1 shown, the recommendation method includes:

[0044] S101: Input the usage information and user profile of the target user within the current reference period into the large model, and obtain multiple interest tags of the target user output by the large model and the confidence of each interest tag.

[0045] In the embodiment of the present disclosure, the target user can be any visiting user in the recommendation system, which may be a heavy user who often uses the recommendation system, or a light user with low usage frequency and low activity.

[0046] In the embodiment of the present disclosure, the reference period is the time period for obtaining information used to predict the user's interest points. To ensure the timeliness of information and the accuracy of prediction, and to avoid the interference of historical information that is too long ago from the current time on the prediction result, the reference period should be the most recent period of time from the current access time of the target user. The length of the reference period can be determined according to the usage frequency of the target user. For example, for a heavy user with a high usage frequency, the reference period can be the most recent week, etc. For a light user with a particularly low usage frequency and a long interval since the last historical visit, the reference period can be a period of time after the current access time.

[0047] In the embodiment of the present disclosure, the usage information of the target user may include the titles of pictures, texts, or videos read in the past. The user profile of the target user may include information such as age, gender, and education level. By analyzing the usage information and user profile of the user, the user's interest points can be determined, and then relevant content can be recommended according to the user's interest points, which is beneficial to improving the user retention rate.

[0048] In the embodiment of the present disclosure, the usage information and user profile can be input into the large model. The large model converts the usage information and user profile into high-dimensional vectors (Embeddings) respectively, and then performs feature encoding through different encoders to form a hidden layer network encoding representation. After being deconstructed by a deep learning network (Transformer), multiple interest tags corresponding to the target user and the confidence of each interest tag can be obtained. The confidence is the degree of certainty of the model for the prediction result. The higher the confidence, the more certain the model is that the tag conforms to the user's interest.

[0049] It should be noted that the large model can be any large language model capable of performing semantic analysis, feature extraction, and feature ranking, such as Wenxin Yiyan, Llama, or Qwen, etc.

[0050] For example, the multiple interest tags of the target user output by the large model are food exploration, anti-Japanese war dramas, and self-driving tours respectively. The confidence level of food exploration is 0.9, the confidence level of anti-Japanese war dramas is 0.85, and the confidence level of self-driving tours is 0.6.

[0051] S102: Obtain reference content from the candidate content library according to the correlation degree between each candidate content in the candidate content library and the interest tags.

[0052] Among them, the candidate content library refers to a database that contains all the graphic and video content that can be recommended and published in the recommendation system.

[0053] In the embodiments of the present disclosure, the correlation degree between each candidate content in the candidate content library and the interest tags can be determined through semantic matching. The higher the degree of semantic matching, the greater the correlation degree. Then, multiple candidate contents can be selected according to the size of the correlation degree as the reference content to be recommended to the target user.

[0054] It should be noted that in the present disclosure, candidate contents with a correlation degree greater than a certain value can be obtained from the candidate content library as reference content, or appropriate numbers of candidate contents with a relatively high correlation degree can also be obtained according to the confidence levels of the associated interest tags as reference content.

[0055] S103: Determine multiple first recommended contents from all the reference contents based on the correlation degree corresponding to each reference content and the confidence level of the associated interest tags.

[0056] In the embodiments of the present disclosure, the correlation degree corresponding to each reference content and the confidence level of the associated interest tags can be comprehensively considered to sort all the reference contents. Those with a higher correlation degree are ranked higher. In the case of the same correlation degree, the greater the confidence level of the associated interest tags, the higher the ranking. Or, based on the correlation degree and the confidence level, the product or the weighted sum value can be calculated, and then sorted according to the size of the value. Then, multiple reference contents can be determined as the first recommended contents according to the order. The multiple first recommended contents form a recommendation queue according to the ranking and are recommended in order during recommendation.

[0057] In the embodiments of the present disclosure, the correlation degree corresponding to each reference content and the confidence level of the associated interest tags can be weighted and summed to obtain the recommendation weight of the reference content. Then, according to the order from large to small of the recommendation weights, the corresponding reference contents are sequentially determined as the first recommended contents, so that the reliability of the recommended content can be determined based on the interest tags of the target user.

[0058] It should be noted that the number of the first recommended content to be determined can be a fixed value, which is determined according to experience; or it can be a variable, such as the number of reference contents whose recommendation weights meet the recommendation conditions (such as being greater than a certain threshold, etc.). The number of the first recommended content to be determined is the same as the number of reference contents that meet the recommendation conditions.

[0059] S104: Determine the second recommended content based on the current recommendation goal of the system.

[0060] In the embodiments of the present disclosure, the system is a recommendation system that applies the recommendation method provided by the present disclosure. The recommendation goal of the system can be click-through rate, conversion rate, etc. For different recommendation goals, different recommended contents can be determined. For example, when the current recommendation goal of the system is the click-through rate, the second recommended content can be the content related to the current hottest topic; or when the current recommendation goal of the system is the conversion rate, the second recommended content can be product promotion, etc. The system can determine different recommendation goals according to the business requirements at different times.

[0061] It should be noted that the second recommended content is the recommended content determined based on the business requirements or goals of the system, without considering the user's interests. The determination methods of the first recommended content and the second recommended content are different, but in some cases, the first recommended content may be the same as the second recommended content.

[0062] S105: Send the second recommended content and at least one first recommended content to the client of the target user.

[0063] Among them, the client of the target user is the browser or application through which the target user accesses the recommendation system. The client can be in any terminal device, such as a personal computer, a smart phone, a tablet computer, etc.

[0064] In the embodiments of the present disclosure, after determining the first recommended content corresponding to the interest tag and the second recommended content corresponding to the system goal, the second recommended content and at least one first recommended content can be sent to the client of the target user for display at the same time. The number of the recommended contents to be sent can be determined according to the number of display slots included in the client interface. The number of display slots of different clients is not necessarily the same.

[0065] It should be noted that in the present disclosure, the first content to be recommended and the second content to be recommended are two different types of recommended content, which are independent of each other. There will be no situation where the second content to be recommended is more popular than the first content to be recommended and is therefore displayed first. This avoids the problem that due to the multi-objective fusion of the recommendation system, other support queues (such as diversified monetization, etc.) may snatch the display slots of content related to the user's interests. The content to be recommended sent to the client must include at least one first content to be recommended, which can, while achieving the system's recommendation goals, ensure that the content recommended to the user each time must include content related to the user's interests, thus avoiding the problem of low retention rate of light users for the recommended content.

[0066] In this embodiment, by obtaining the interest tags and confidence levels based on the user usage information and portrait analysis, obtaining the content to be recommended related to the interests, and obtaining the content to be recommended that meets the system's recommendation goals, the content to be recommended determined by the two methods is sent to the client for display at the same time. The content to be recommended sent includes at least one content to be recommended related to the interests, which can, while achieving the system's recommendation goals, ensure that the content recommended to the user each time must include content related to the user's interests, improving the reliability and accuracy of content recommendation and being beneficial to improving the satisfaction and retention rate of users, especially light users.

[0067] Figure 2 It is a schematic flowchart of a recommendation method proposed in another embodiment of the present disclosure.

[0068] As Figure 2 shown, the recommendation method includes:

[0069] S201: Input the usage information and user portrait of the target user in the current reference period into the large model to obtain multiple interest tags of the target user output by the large model and the confidence level of each interest tag.

[0070] For the description of the above S201, specific reference can be made to other embodiments of the present disclosure, which will not be elaborated here.

[0071] S202: Based on the confidence level of each interest tag, determine the number of content to be recalled corresponding to the interest tag.

[0072] In the embodiments of the present disclosure, the higher the confidence level of the interest tag, the more the number of content to be recalled corresponding to the interest tag. Because when the confidence level of a certain interest tag is high, it can be determined that the tag more conforms to the current interested tendency of the target user, and then more relevant recalled content can be determined based on the interest tag for recommendation, further improving the accuracy of interest prediction-based recommendation.

[0073] S203: For each interest tag, obtain the number of reference contents to be recalled corresponding to it from the candidate content library according to the degree of association between each candidate content and the interest tag.

[0074] In the embodiments of the present disclosure, for each interest tag, in the candidate knowledge base, the degree of association between each candidate content and the interest tag can be obtained by calculating the semantic similarity between each candidate content and the interest tag. Then, all candidate contents can be sorted according to the magnitude of the degree of association, and the number of candidate contents corresponding to the interest tag to be recalled can be obtained in order as the reference contents corresponding to the interest tag.

[0075] It should be noted that in some embodiments, the same candidate content may be recalled for different interest tags. In this case, the candidate content is used as a reference content, corresponding to multiple degrees of association and the confidence levels of the associated interest tags. When performing recommendation ranking on the reference content later, it is necessary to consider the degree of association and confidence level with each associated interest tag to ensure the reliability of content recommendation.

[0076] In this embodiment, when obtaining the associated recommended content for the interest tag according to the confidence level of the interest tag, controlling the number of recalled contents can ensure a greater probability of recommending the content that the user is more interested in, and further improve the accuracy of interest prediction-based recommendation.

[0077] S204: Determine multiple first contents to be recommended from all reference contents based on the degree of association corresponding to each reference content and the confidence level of the associated interest tag.

[0078] S205: Determine the second content to be recommended based on the current recommendation goal of the system.

[0079] S206: Send the second content to be recommended and at least one first content to be recommended to the client of the target user.

[0080] For the descriptions of S204 to S206 above, specific references can be made to other embodiments of the present disclosure, which will not be elaborated here.

[0081] Figure 3 It is a schematic flowchart of a recommendation method proposed in another embodiment of the present disclosure.

[0082] As Figure 3 shown, the recommendation method includes:

[0083] S301: Input the usage information and user portrait of the target user in the current reference period into the large model to obtain multiple interest tags of the target user output by the large model and the confidence level of each interest tag.

[0084] S302: Obtain reference content from the candidate content library according to the association degree between each candidate content and the interest tags in the candidate content library.

[0085] For the descriptions of S301 and S302 above, specific references can be made to other embodiments of the present disclosure, which will not be elaborated here.

[0086] S303: For each reference content, determine the recommendation weight of the reference content according to its corresponding association degree and the confidence level of the associated interest tags.

[0087] In the embodiments of the present disclosure, calculations such as multiplication or weighted summation can be performed on the association degree corresponding to each reference content and the confidence level of the associated interest tags to obtain the recommendation weight of the reference content.

[0088] In the embodiments of the present disclosure, there may be a situation where the same reference content is associated with multiple interest tags, and the confidence levels of the multiple interest tags are different, and the association degrees with each interest tag are also different. At this time, the corresponding association degrees can be weighted and summed based on the confidence level of each interest tag, and the value after weighted summation is used as the recommendation weight of the reference content.

[0089] For example, the interest tags associated with any reference content are tag A and tag B, the association degree of the reference content with tag A is 0.5, the confidence level of tag A is 0.8, the association degree of the reference content with tag B is 0.7, and the confidence level of tag B is 0.6. Then the recommendation weight of the reference content is 0.8 * 0.5 + 0.7 * 0.6 = 0.82.

[0090] S304: Based on the recommendation weights, sort all the reference contents to determine multiple first to-be-recommended contents whose weight values are greater than the weight threshold, or determine at least m first to-be-recommended contents with the largest recommendation weight values.

[0091] Among them, the weight value is the value of the recommendation weight of the reference content, and the weight threshold can be determined according to the accuracy requirements of the actual recommendation system. The present disclosure does not make any limitations in this regard.

[0092] Among them, m is an integer greater than 1, and the value of m can also be determined according to accuracy requirements, etc.

[0093] For example, all the reference contents include content a, content b, content c, and content d, and the corresponding recommendation weights are 0.82, 0.7, 0.65, and 0.4 respectively. When the weight threshold is 0.6, since 0.82 > 0.7 > 0.65 > 0.6 > 0.4, the first to-be-recommended contents are content a, content b, and content c. Or, when the number m of the first to-be-recommended contents is 2, according to the sorting result of the weight values 0.82 > 0.7 > 0.65 > 0.4, the first to-be-recommended contents are content a and content b.

[0094] In this embodiment, by using the correlation degree and confidence level, the recommendation weight of each reference content is calculated, and then based on the magnitude and ranking of the recommendation weights, some qualified reference contents are selected as the contents to be recommended, so that the reliability and accuracy of the recommended contents can be improved.

[0095] S305: Determine the second content to be recommended based on the current recommendation target of the system.

[0096] S306: Send the second content to be recommended and at least one first content to be recommended to the client of the target user.

[0097] For the descriptions of S305 and S306 above, specific references can be made to other embodiments of the present disclosure, which will not be elaborated here.

[0098] It should be noted that after sending the second content to be recommended and at least one first content to be recommended to the client in the above embodiment, the user may need to refresh the recommended content in the client interface due to reasons such as finishing browsing or not being interested, in order to obtain new recommended content. Therefore, the recommendation method proposed by the present disclosure may further include the process of refreshing the recommended content.

[0099] Figure 4 It is a schematic flowchart of a recommendation method proposed by another embodiment of the present disclosure.

[0100] As Figure 4 shown, the recommendation method includes:

[0101] S401: In the case of receiving a recommendation content refresh request sent by the client, determine the first interest tag associated with the first content to be recommended currently displayed on the client.

[0102] In the embodiment of the present disclosure, after sending the second content to be recommended and the first content to be recommended to the client, operations such as the user swiping down the page or clicking the refresh control can trigger the client to send a recommendation content refresh request to the recommendation system. Then, after receiving the request, the recommendation system can first determine the first interest tag associated with the first content to be recommended currently displayed, so as to determine the first content to be recommended for the next display after the refresh.

[0103] S402: Select at least one first content to be recommended associated with the second interest tag from multiple first contents to be recommended.

[0104] Among them, the second interest tag can be a tag different from the first interest tag among all the interest tags predicted for the target user, or the second interest tag can also be the same as the first interest tag.

[0105] In the embodiments of the present disclosure, in order to determine the first interest tag associated with the first content to be recommended currently displayed on the client and for the diversity of the recommended content, one or more second interest tags different from the first interest tag may be randomly determined from the multiple interest tags output by the large model for the target user. Then, at least one first content to be recommended associated with the second interest tag can be obtained from the multiple first contents to be recommended determined based on the degree of association and confidence.

[0106] It should be noted that in the present disclosure, multiple first contents to be recommended can be displayed in turn according to the size of the degree of association and confidence. Therefore, after receiving a request to refresh the recommended content, the recommendation system can obtain at least one content to be recommended after the first content to be recommended currently displayed from the queue of multiple first contents to be recommended. Since the queue of the first contents to be recommended is sorted according to the degree of association and confidence, it is possible that the second interest tag associated with the newly obtained first content to be recommended is the same as the first interest tag associated with the first content to be recommended currently displayed.

[0107] In the embodiments of the present disclosure, the number of first contents to be recommended sent to the client for display each time can be the same or different.

[0108] S403: Determine the current second content to be recommended.

[0109] In the embodiments of the present disclosure, the current second content to be recommended is the content different from the second content to be recommended currently displayed on the client among all the second contents to be recommended determined based on the current recommendation target of the system.

[0110] S404: Return the current second content to be recommended and at least one first content to be recommended associated with the second interest tag to the client.

[0111] In some possible embodiments, there may be a situation where the user swipes down the page but does not completely refresh the currently displayed recommended content. At this time, the number of second contents to be recommended and first contents to be recommended returned to the client can be determined according to the length of the swiped page.

[0112] In this embodiment, after sending the second content to be recommended and at least one of the first contents to be recommended to the client of the target user, when a request to refresh the recommended content is received, the content to be recommended corresponding to the new interest tag and the content to be recommended corresponding to the system target are obtained and returned to the client for updated display, implementing a forced exposure and rotation mechanism for user interest-related content, improving the stickiness between the recommended content and the user, and improving the diversity of the recommended content, thus optimizing the user experience.

[0113] It should be noted that for different types of clients (such as mobile phones, computers, etc.), the number of contents that can be displayed each time is different. Each content has its corresponding display slot, and according to actual needs, the ratio of personal recommendations to system recommendations in the display slots may also be different. For example, during a specific festival, more content related to activities needs to be displayed in the client display slots, then the slots for user personal interest recommendations may be reduced accordingly, but there must be at least one slot. Therefore, when sending the second content to be recommended and the first content to be recommended to the client, it can be determined according to the number of available display slots in the client.

[0114] Figure 5 It is a schematic flowchart of a recommendation method proposed in another embodiment of the present disclosure.

[0115] As Figure 5 shown, the recommendation method includes:

[0116] S501: Input the usage information and user portrait of the target user during the current reference period into the large model to obtain multiple interest tags of the target user output by the large model and the confidence of each interest tag.

[0117] S502: Obtain reference content from the candidate content library according to the correlation between each candidate content in the candidate content library and the interest tags.

[0118] S503: Determine multiple first contents to be recommended from all the reference contents based on the correlation corresponding to each reference content and the confidence of the associated interest tags.

[0119] S504: Determine the second content to be recommended based on the current recommendation goal of the system.

[0120] For the descriptions of the above steps S501 to S504, specific references can be made to other embodiments of the present disclosure, which will not be elaborated here.

[0121] S505: Determine the number of currently available display slots in the client of the target user.

[0122] Among them, the display slot is the position for content display in the client interface, and each recommended content corresponds to a display slot. The number of currently available display slots refers to the number of slots that can be used for recommending content related to user interests except for the display slots occupied by system-recommended content according to the current business needs of the client. The number of currently available display slots must be greater than or equal to 1.

[0123] In the embodiments of the present disclosure, the total number of display slots in the client interface can be determined according to the type of the client of the target user. Each type of client may include a different number of display slots according to the interface design. For example, the number of display slots in the mobile client interface can be 6 or 8, etc. The currently available number of display slots should be less than the total number of display slots in the client interface and greater than or equal to 1, and can be determined according to the current business needs of the recommendation system.

[0124] Alternatively, in some possible embodiments, there may be a situation where the user swipes down the page but does not fully refresh the currently displayed recommended content. At this time, the currently available number of display slots in the client can be determined according to the length of the swiped page.

[0125] S506: Determine the number n of the first to-be-recommended contents to be returned according to the number of display slots.

[0126] In the embodiments of the present disclosure, the number of available display slots determines the number of the first to-be-recommended contents to be returned.

[0127] S507: Send the second to-be-recommended content and n first to-be-recommended contents to the client of the target user.

[0128] Optionally, when n is greater than 1, n first to-be-recommended contents respectively associated with different interest tags can be obtained from multiple first to-be-recommended contents, and then the second to-be-recommended content and n first to-be-recommended contents respectively associated with different interest tags are sent to the client of the target user. Thus, it can be ensured that when recommending multiple interest-related contents to the user, the probability that the contents meet the user's consumption needs is greater, the diversity of content recommendation is improved, and it is more conducive to improving the user retention rate and optimizing the user experience.

[0129] In this embodiment, by determining the currently available number of display slots in the client, obtaining the corresponding number of interest-related recommended contents, and returning them to the user client for display, the recommendation efficiency and performance can be improved.

[0130] Optionally, in the present disclosure, when the target user is a light user of the system, the second to-be-recommended content and at least one of the first to-be-recommended contents are sent to the client of the target user.

[0131] Among them, a light user refers to a user who has few access times and low frequency to the recommendation system.

[0132] Since when recommending contents to light users, it is easily affected by the popularity of the mainstream population in the market (such as male population, online money-making population), it is easy to recommend high-heat but irrelevant contents to the users, which is not conducive to the retention of light users.

[0133] Therefore, the present invention can ensure the stickiness of the recommended content and the user while achieving the system recommendation goal by forcing the content to be sent to include at least one first content to be recommended related to the user's interests when sending recommended content to light users, avoiding the problem of low retention rate of recommended content for light users, which is conducive to improving the retention rate of light users.

[0134] Optionally, the number k of first contents to be recommended to be returned may be determined based on the frequency, duration and interaction information of the target user using the system, and then the second contents to be recommended and k first contents to be recommended are sent to the client of the target user.

[0135] The interest tags associated with the k first contents to be recommended are different or not completely the same.

[0136] In the disclosed embodiment, the viscosity between the user and the recommendation system can be determined based on the frequency, duration and interaction information of the target user's use of the system. The higher the frequency of use, the longer the time and the more interaction information, the heavier the viscosity between the user and the recommendation system, and vice versa. The lighter the user, the more first content to be recommended can be returned, so that the user's viscosity can be increased by sending recommended content to the user client, and the conversion rate can be considered after the viscosity between the user and the recommendation system reaches a certain level.

[0137] In the disclosed embodiment, by determining the amount of interest-related content recommended to the user each time based on the target user's frequency, duration, and interaction information of using the system, the proportion of recommended content can be adjusted in a targeted manner, thereby optimizing the user experience of light users while ensuring the conversion rate of the entire system, which is conducive to improving user retention rate and enhancing the competitiveness of the recommendation system.

[0138] Figure 6 It is a structural diagram of a recommended device proposed in an embodiment of the present disclosure.

[0139] like Figure 6 As shown, the recommendation device 60 comprises:

[0140] Processing module 601, used to input the target user's usage information and user portrait in the current reference period into the big model, and obtain multiple interest tags of the target user and the confidence of each interest tag output by the big model;

[0141] An acquisition module 602 is used to acquire reference content from the candidate content library according to the correlation between each candidate content and the interest tag in the candidate content library;

[0142] The first determination module 603 is configured to determine a plurality of first content to be recommended from all reference contents based on the relevance corresponding to each reference content and the confidence of the associated interest tags;

[0143] The second determination module 604 is configured to determine second content to be recommended based on the current recommendation target of the system;

[0144] The sending module 605 is configured to send the second content to be recommended and at least one first content to be recommended to the client of the target user.

[0145] In some embodiments, the obtaining module 602 may specifically be configured to:

[0146] Determine the number of contents to be recalled corresponding to each interest tag based on the confidence of each interest tag;

[0147] For each interest tag, obtain the number of reference contents corresponding to it from the candidate content library according to the relevance between each candidate content and the interest tag.

[0148] In some embodiments, the first determination module 603 may specifically be configured to:

[0149] For each reference content, determine the recommendation weight of the reference content according to the corresponding relevance and the confidence of the associated interest tag;

[0150] Based on the recommendation weight, sort all reference contents to determine a plurality of first contents to be recommended with a weight value greater than the weight threshold, or determine at least m first contents to be recommended with the largest recommendation weight values, where m is an integer greater than 1.

[0151] In some embodiments, the sending module 605 may further be configured to:

[0152] In the case of receiving a recommendation content refresh request sent by the client, determine the first interest tag associated with the first content to be recommended currently displayed on the client;

[0153] Select at least one first content to be recommended associated with a second interest tag from the plurality of first contents to be recommended;

[0154] Determine the current second content to be recommended;

[0155] Return the current second content to be recommended and at least one first content to be recommended associated with the second interest tag to the client.

[0156] In some embodiments, the sending module 605 may specifically be configured to:

[0157] Determine the number of currently available display slots in the client of the target user;

[0158] Determine the number n of the first recommended contents to be returned according to the number of display slots;

[0159] Send the second recommended content and n first recommended contents to the client of the target user.

[0160] In some embodiments, the sending module 605 may specifically be configured to:

[0161] Obtain n first recommended contents respectively associated with different interest tags from multiple first recommended contents;

[0162] Send the second recommended content and n first recommended contents respectively associated with different interest tags to the client of the target user.

[0163] In some embodiments, the sending module 605 may specifically be configured to:

[0164] In the case that the target user is a light user of the system, send the second recommended content and at least one first recommended content to the client of the target user.

[0165] In some embodiments, the sending module 605 may specifically be configured to:

[0166] Determine the number k of the first recommended contents to be returned according to the frequency, duration, and interaction information of the target user using the system;

[0167] Send the second recommended content and k first recommended contents to the client of the target user, where the k first recommended contents are respectively associated with different or not completely the same interest tags.

[0168] It should be noted that the foregoing explanation of the recommendation method also applies to the recommendation device in this embodiment, and will not be elaborated herein.

[0169] In this embodiment, by obtaining the recommended contents related to interests and the recommended contents meeting the system recommendation goals according to the interest tags and confidence levels obtained from the user usage information and portrait analysis, and sending the recommended contents determined by the two methods to the client for display at the same time, the sent recommended contents include at least one recommended content related to interests, which can ensure that each content recommended to the user must include the content related to the user's interests while achieving the system recommendation goal, improving the reliability and accuracy of content recommendation, and being beneficial to improving the satisfaction and retention rate of users, especially light users.

[0170] According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0171] Figure 7 FIG. Figure 7 shows a schematic block diagram of an exemplary electronic device 700 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, for example, 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, for example, 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 exemplary and are not intended to limit the implementations of the present disclosure described and / or claimed herein.

[0172] As Figure 7 shown, the device 700 includes a computing unit 701 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0173] A plurality of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0174] The computing unit 701 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing 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. The computing unit 701 executes the various methods and processes described above, such as the recommendation method. For example, in some embodiments, the recommendation method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the recommendation method described above can be executed. Alternatively, in other embodiments, the computing unit 701 can be configured to execute the recommendation method in any other suitable way (e.g., by means of firmware).

[0175] Various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SOCs), 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 can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor, and can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0176] 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, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program codes can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0177] In the context of this 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.

[0178] 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).

[0179] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (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 a computing system that includes any combination of such back-end components, middleware components, 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: local area network (LAN), wide area network (WAN), the Internet, and blockchain network.

[0180] A computer system may include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services ("Virtual Private Server", or simply "VPS" for short). The server may also be a server of a distributed system or a server combined with a blockchain.

[0181] It should be understood that various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and no limitation is imposed herein.

[0182] In addition, the terms "first" and "second" are used only for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of this disclosure, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined. In the description of this disclosure, the words "if" and "when" can be interpreted as "when...", "while...", "in response to determining", or "in... case".

[0183] 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 recommendation method, comprising: Inputting the usage information and user profile of the target user within the current reference period into a large model to obtain multiple interest tags of the target user output by the large model and the confidence level of each interest tag; Obtaining reference content from the candidate content library according to the correlation degree between each candidate content in the candidate content library and the interest tags; Determining multiple first content to be recommended from all the reference content based on the correlation degree corresponding to each reference content and the confidence level of the associated interest tags; Determining second content to be recommended based on the current recommendation goal of the system; Sending the second content to be recommended and at least one of the first content to be recommended to the client of the target user.

2. The method according to claim 1, wherein The obtaining reference content from the candidate content library according to the correlation degree between each candidate content in the candidate content library and the interest tags includes: Determining the number of content to be recalled corresponding to each interest tag based on the confidence level of each interest tag; For each interest tag, obtaining the number of reference content corresponding to it from the candidate content library according to the correlation degree between each candidate content and the interest tag.

3. The method according to claim 1, wherein, The determining multiple first content to be recommended from all the reference content based on the correlation degree corresponding to each reference content and the confidence level of the associated interest tags includes: For each reference content, determining the recommendation weight of the reference content according to its corresponding correlation degree and the confidence level of the associated interest tag; Based on the recommendation weight, sorting all the reference content to determine multiple first content to be recommended with a weight value greater than the weight threshold, or determining at least m first content to be recommended with the largest recommendation weight values, where m is an integer greater than 1.

4. The method according to claim 1, wherein After the sending the second content to be recommended and at least one of the first content to be recommended to the client of the target user, it further includes: When receiving a recommendation content refresh request sent by the client, determining the first interest tag associated with the first content to be recommended currently displayed on the client; Selecting at least one first content to be recommended associated with a second interest tag from the multiple first content to be recommended; Determining the current second content to be recommended; Returning the current second content to be recommended and the at least one first content to be recommended associated with the second interest tag to the client.

5. The method according to any one of claims 1-4, wherein, The sending the second content to be recommended and at least one of the first content to be recommended to the client of the target user includes: Determining the number of available display slots currently in the client of the target user; Determining the number n of the first content to be recommended to be returned according to the number of display slots; Sending the second content to be recommended and n of the first content to be recommended to the client of the target user.

6. The method according to claim 5, wherein, The n is a value greater than 1, and the sending the second content to be recommended and n of the first content to be recommended to the client of the target user includes: Obtaining n first content to be recommended respectively associated with different interest tags from the multiple first content to be recommended; Send the second content to be recommended and the n first contents to be recommended respectively associated with different interest tags to the client of the target user.

7. The method according to any one of claims 1-4, wherein, The sending the second content to be recommended and at least one of the first contents to be recommended to the client of the target user includes: In the case where the target user is a light user of the system, send the second content to be recommended and at least one of the first contents to be recommended to the client of the target user.

8. The method according to any one of claims 1-4, wherein, The sending the second content to be recommended and at least one of the first contents to be recommended to the client of the target user includes: Determine the number k of the first contents to be recommended to be returned according to the frequency, duration and interaction information of the target user using the system; Send the second content to be recommended and k of the first contents to be recommended to the client of the target user, where the k first contents to be recommended are respectively associated with different or not completely the same interest tags.

9. A recommendation device, comprising: A processing module, configured to input the usage information and user portrait of a target user in a current reference period into a large model, and obtain multiple interest tags of the target user and the confidence level of each interest tag output by the large model; An acquisition module, configured to acquire reference content from the candidate content library according to the association degree between each candidate content in the candidate content library and the interest tags; A first determination module, configured to determine multiple first contents to be recommended from all the reference contents based on the association degree corresponding to each reference content and the confidence level of the associated interest tag; A second determination module, configured to determine a second content to be recommended based on the current recommendation target of the system; A sending module, configured to send the second content to be recommended and at least one of the first contents to be recommended to the client of the target user.

10. An electronic device, comprising: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the recommendation method according to any one of claims 1-8.

11. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, Wherein, The computer instructions are used to cause the computer to execute the recommendation method according to any one of claims 1-8.

12. A computer program product, characterized in that, Including a computer program, which when executed by a processor, implements the steps of the recommendation method according to any one of claims 1-8.

Images