Content recommendation data processing method and device based on large model, equipment and medium

Abstract

The invention provides a content recommendation data processing method and device based on a large model, equipment and a medium, and relates to the field of artificial intelligence, in particular to the field of content recommendation. According to the specific implementation scheme, user information of a to-be-recommended user is obtained; wherein the user information comprises a user portrait and historical consumption information in a historical time period; determining potential consumption content based on a preset large model according to the user information; and according to the user information and the potential consumption content, determining to-be-recommended content, and pushing the to-be-recommended content to the to-be-recommended user. According to the user portrait and the known historical consumption information, the potential consumption content which may be consumed by the user is speculated, so that the consumption performance of the user is expanded, and the content recommendation precision is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of content recommendation in the field of artificial intelligence, and in particular to a content recommendation data processing method, device, equipment and medium based on a large model. Background Art

[0002] When people use APP (Application) to watch content, they usually hope that the APP can actively push content that they may be interested in, thereby reducing their own search operations.

[0003] If the content pushed by the APP is not what the user is interested in, it will reduce the user experience and is not conducive to enhancing the stickiness between the user and the APP product. Therefore, it is necessary to recommend accurate content to users to improve the user experience. Summary of the invention

[0004] The present invention provides a content recommendation data processing method, device, equipment and medium based on a large model.

[0005] According to a first aspect of the present disclosure, a content recommendation data processing method based on a large model is provided, comprising:

[0006] Obtaining user information of the user to be recommended; wherein the user information includes user portrait and historical consumption information within a historical time period;

[0007] According to the user information, based on a preset macro model, determine potential consumption content;

[0008] Determining content to be recommended based on the user information and the potential consumption content;

[0009] The content to be recommended is pushed to the user to be recommended.

[0010] According to a second aspect of the present disclosure, a content recommendation data processing device based on a large model is provided, comprising:

[0011] An information acquisition unit, used to acquire user information of the user to be recommended; wherein the user information includes a user portrait and historical consumption information within a historical time period;

[0012] A content determination unit, configured to determine potential consumption content according to the user information and based on a preset macro model;

[0013] A recommendation determination unit, configured to determine content to be recommended based on the user information and the potential consumption content;

[0014] The content push unit is used to push the content to be recommended to the user to be recommended.

[0015] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0016] at least one processor; and

[0017] a memory communicatively coupled to the at least one processor;

[0018] The memory stores instructions that can be executed 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 perform the method described in the first aspect of the present disclosure.

[0019] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method according to the first aspect of the present disclosure.

[0020] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the steps of the method described in the first aspect of the present disclosure when executed by a processor.

[0021] According to the technology disclosed in the present invention, the accuracy of content recommendation for users is improved.

[0022] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended 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

[0023] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0024] Figure 1 It is a flowchart of a content recommendation data processing method based on a large model provided according to an embodiment of the present disclosure;

[0025] Figure 2 It is a flowchart of a content recommendation data processing method based on a large model provided according to an embodiment of the present disclosure;

[0026] Figure 3 It is a flowchart of a content recommendation data processing method based on a large model provided according to an embodiment of the present disclosure;

[0027] Figure 4 is a schematic diagram of a content recommendation process provided according to an embodiment of the present disclosure;

[0028] Figure 5 It is a structural block diagram of a content recommendation data processing device based on a large model provided according to an embodiment of the present disclosure;

[0029] Figure 6 It is a structural block diagram of a content recommendation data processing device based on a large model provided according to an embodiment of the present disclosure;

[0030] Figure 7 is a block diagram of an electronic device for implementing a content recommendation data processing method based on a large model according to an embodiment of the present disclosure;

[0031] Figure 8 It is a block diagram of an electronic device used to implement a content recommendation data processing method based on a large model according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0032] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0033] Users can consume content through APP products. For example, users can click or browse pictures, texts, videos, etc. As users use the APP, the APP records the user's content consumption preferences, so that it can recommend content to users and actively push content that users may be interested in to users, thereby improving user experience.

[0034] However, for new and light users of the APP, there are not many consumption records recorded in the APP, so it is impossible to grasp the user's consumption preferences and recommend content to the user. At present, for new and light users, recommendations are usually made based on the consumption of the general market users. This recommendation method is affected by the popularity of the general market, which reduces the degree of personalization of the recommendation and cannot make accurate content recommendations for new and light users, resulting in a decline in user experience.

[0035] The present invention provides a content recommendation data processing method, device, equipment and storage medium based on a large model, which are applied to the content recommendation field in the field of artificial intelligence to improve the accuracy of content recommendation.

[0036] It should be noted that the model in this embodiment is not a model for a specific user and cannot reflect the relevant information of a specific user. It should be noted that the data in this embodiment comes from a public data set.

[0037] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0038] In order to enable readers to more deeply understand the implementation principle of the present disclosure, the following Figure 1-Figure 8 The embodiment is further refined.

[0039] Figure 1 The flowchart of a method for processing content recommendation data based on a large model provided according to an embodiment of the present disclosure is shown in FIG. 1 , which can be executed by a device for processing content recommendation data based on a large model. Figure 1 As shown, the method comprises the following steps:

[0040] S101. Obtain user information of the user to be recommended; wherein the user information includes user portrait and historical consumption information within a historical time period.

[0041] For example, users can view content in the form of pictures, texts, videos, etc. through various software products. For example, there are many video browsing apps, and users can watch different videos by clicking or sliding. Users' actions such as clicking, browsing, liking, commenting, collecting, and forwarding content can be regarded as users' consumption behaviors, and users' consumption behaviors can be recorded and stored in real time. For example, the time when users browse videos, the titles and authors of the browsed videos, etc. can be stored.

[0042] The user can also record relevant information in the software product. For example, the user can enter his / her age, gender and other information into the software product. Such relevant information of the user constitutes the user profile of the user. In this embodiment, before obtaining the user information, the user knows that the information will be obtained and the user agrees.

[0043] When a user consumes content using a software product, the software product can actively recommend content to the user. The user can be determined as a user to be recommended. When recommending content, user information of the user to be recommended can be obtained. User information can include user portraits and historical consumption information. The user portrait represents relevant information about the user, and the historical consumption information represents the consumption of the user in a preset historical time period.

[0044] For new and light users, a small amount of historical consumption information can be obtained. New users refer to newly registered users, and light users refer to users who consume content less frequently. These two types of users have less consumption behavior. If recommendations are made based only on user portraits and historical consumption information, it may not be possible to make accurate recommendations to users, affecting user experience.

[0045] S102: Determine potential consumption content based on user information.

[0046] Exemplarily, after obtaining the user information, the potential consumption content of the user to be recommended can be determined based on the user portrait and historical consumption information in the user information. Potential consumption content represents the content that the user to be recommended may consume, which is inferred from the preset historical content. The preset historical content refers to the content published in the software product in the past period of time, for example, it can be content uploaded by other users, and the user to be recommended has not consumed these contents. For example, the user to be recommended has not logged into the APP in the past period of time, or has not registered an account, resulting in no consumption behavior. The time period corresponding to the preset historical content can be greater than the time period corresponding to the historical consumption information, that is, the time period corresponding to the preset historical content can be longer. It is inferred which content the user to be recommended may consume if he has logged into the APP in the past period of time, and the inferred content is determined as potential consumption content.

[0047] Based on the user portrait, the tags of interest to the recommended user can be determined. The tags of interest can be tags selected or set by the user, indicating the areas of interest to the user. Each content can correspond to one or more areas, and the content in the areas consistent with the tags of interest can be found from the preset historical content as potential consumption content. It is also possible to find the content of the same author from the preset historical content based on the historical consumption information as potential consumption content. It is also possible to compare the preset historical content with the content in the historical consumption information, and select the historical content with high similarity as the potential consumption content.

[0048] In this embodiment, the user information includes historical consumption, and the historical consumption represents the amount of content consumed by the to-be-recommended user within a preset historical time period; the method also includes: in response to the historical consumption being less than or equal to a preset quantity threshold, determining the potential consumption content of the to-be-recommended user based on the user portrait and historical consumption information; determining the content to be recommended based on the user portrait, historical consumption information and potential consumption content, and pushing the content to be recommended to the to-be-recommended user; in response to the historical consumption being greater than a preset quantity threshold, determining the content to be recommended based on the user portrait and historical consumption information, and pushing the content to be recommended to the to-be-recommended user.

[0049] Specifically, the user information may also include historical consumption, which represents the amount of content consumed by the to-be-recommended user within a preset historical time period, that is, how much content the user has consumed.

[0050] A quantity threshold is set in advance, and the historical consumption is compared with the preset quantity threshold. If the historical consumption is less than or equal to the preset quantity threshold, it means that the user has not consumed much content, and the content that has been consumed is not enough to support accurate content recommendation for the user. S102 and S103 in this embodiment can be continued, that is, based on the user portrait and historical consumption information, the potential consumption content of the user to be recommended is determined, and then based on the user portrait, historical consumption information and potential consumption content, the content to be recommended is determined, and the content to be recommended is pushed to the user to be recommended; if the historical consumption is greater than the preset quantity threshold, it means that the user has consumed a lot of content, and the content to be recommended can be determined directly based on the user portrait and historical consumption information, without the need to expand the user's performance by determining the potential consumption content.

[0051] The beneficial effect of such a setting is that for new and light users, if the user's historical consumption is insufficient, potential consumption content can be determined to expand user performance, improve the accuracy of determining the content to be recommended, and achieve personalized recommendations.

[0052] S103: Determine content to be recommended based on user information and potential consumption content.

[0053] Exemplarily, after obtaining the potential consumption content, the content to be recommended may be determined by combining the user portrait, historical consumption information, and the potential consumption content.

[0054] The historical consumption information and potential consumption content can be used together as the user's consumption performance to determine what type of content the user likes to consume. Combined with the user's profile, the content that is more in line with the user can be found as the content to be recommended. For example, through historical consumption information and potential consumption content, it can be determined that the user likes to consume content related to animals and plants. If the occupation in the user's profile is an animal keeper, animal-related content can be recommended to the user.

[0055] S104: Push the content to be recommended to the user to be recommended.

[0056] Exemplarily, the content to be recommended may be pushed to the user to be recommended, for example, may be pushed to a terminal device of the user to be recommended.

[0057] In the disclosed embodiment, the user portrait and historical consumption information of the user to be recommended are obtained, and the content that the user may consume in the past large amount of historical content is inferred through the user portrait and known historical consumption information, as potential consumption content, to expand the user's performance. Then, based on the user portrait, as well as the known and inferred consumption content, the content that can be recommended to the user is determined. When the user's historical consumption information is insufficient, the content to be recommended is prevented from being affected by the popularity of the market, thereby realizing personalized recommendation of content, improving the accuracy of content recommendation, and enhancing user experience.

[0058] Figure 2 A flowchart of a content recommendation data processing method based on a large model provided in an embodiment of the present disclosure.

[0059] In this embodiment, potential consumption content is determined based on user information and a preset big model, including: determining interest categories based on user information and a preset big model; wherein the interest categories are related to the categories of historical consumption information; and determining potential consumption content based on the interest categories.

[0060] like Figure 2 As shown, the method comprises the following steps:

[0061] S201. Obtain user information of the user to be recommended; wherein the user information includes a user portrait and historical consumption information within a historical time period.

[0062] Exemplarily, this step may refer to the above-mentioned step S101 and will not be described in detail.

[0063] S202: Determine interest categories according to user information and based on a preset macro model; wherein the interest categories are related to categories of historical consumption information.

[0064] Exemplarily, each content has its own content category, which may refer to the field to which the content belongs. For example, the content category may include food cooking, travel and adventure, culture and history, fashion and beauty, etc. When uploading content, users can set the content category to which the content belongs, and one content may correspond to one or more content categories.

[0065] After obtaining the user portrait and historical consumption information, the interest category of the user to be recommended can be determined. The interest category is the content category that the user is interested in. The content consumed by the user to be recommended in a preset historical time period can be determined first, and the content categories related to these contents can be used as interest categories. That is, the interest category can be the content category of the content consumed historically, or it can be a content category other than the content category of the content consumed historically. For example, the content category of the content consumed historically by the user is the category of fitness tutorials. Fitness tutorials and nutritional diet are both aspects of healthy living. In this case, the categories of fitness tutorials and nutritional diet can both be used as interest categories.

[0066] According to the user portrait, content categories related to the user portrait can also be determined as interest categories. For example, if the user portrait indicates that the user to be recommended is a young woman, the content category of fashion and beauty can be used as the interest category.

[0067] In this embodiment, the interest category of the user to be recommended is determined based on the user information and a preset macro model, including: generating prompt words based on a preset prompt word template according to the user information; inputting the prompt words into the preset macro model to obtain an output interest category.

[0068] Specifically, a template with prompt words is pre-set. The prompt words are text data applied to the large model, which can assist the large model to better understand the task. According to the user portrait and historical consumption information, the preset prompt word template can be supplemented, and the prompt words are obtained after the supplement, that is, the user portrait and historical consumption information can be represented in the prompt words. For example, the preset prompt word template is "You are a short video recommendation system. The user's gender is XXX, age is XXX, and education level is XXX. Mainly consume videos of categories such as XXX. Based on the user portrait and consumption records, what other categories of videos can be recommended?" Among them, "XXX" is the part of the user portrait and historical consumption information that needs to be supplemented.

[0069] A large model, that is, a large language model, is pre-built. Based on the prompt words and the large model, the interest category of the user to be recommended can be obtained. The input data of the large model can be the prompt words, and the output data can be the interest category. In this embodiment, the model structure of the large model is not specifically limited.

[0070] The beneficial effect of such a setting is that, based on user information, personalized prompt words can be generated, thereby improving the accuracy and efficiency of determining interest categories, thereby improving the accuracy and efficiency of content recommendations.

[0071] In this embodiment, the prompt word is input into a preset macro model to obtain an output interest category, including: inputting the prompt word into the macro model to obtain semantic information of the output prompt word; and determining the interest category according to the semantic information.

[0072] Specifically, the generated prompt word is input into a preset big model, and the big model can perform semantic analysis and other processing on the prompt word to obtain the semantic information of the prompt word. According to the semantic information of the prompt word, the interest category of the user to be recommended can be output. For example, the semantic information can be subjected to feature extraction processing by the big model to extract the feature vector, calculate the matching degree between the feature vector and each content category, and determine the content category with the high matching degree as the interest category.

[0073] In this embodiment, the feature extraction process and calculation process within the large model are not specifically limited, as long as the interest category can be extracted from the prompt word.

[0074] The beneficial effect of such a setting is that, by using a large model to perform semantic recognition on prompt words, interest categories are automatically obtained, the efficiency and accuracy of determining interest categories are improved, and intelligent content recommendation is achieved.

[0075] S203: Determine potential consumption content based on interest categories.

[0076] Exemplarily, after obtaining the interest category, content related to the interest category can be searched from the preset historical content as potential consumption content. For example, historical content belonging to the interest category can be determined, and historical content with high browsing volume and belonging to the interest category can be determined as potential consumption content.

[0077] In this embodiment, based on the information that the user has consumed, the categories that the user is interested in are determined, and then the potential consumption content that the user may consume is determined, thereby expanding the user's consumption performance. This achieves information gain, difference, and correlation between the potential consumption content and the historical consumption content, thereby improving the accuracy of content recommendation.

[0078] In this embodiment, determining potential consumption content according to interest categories includes: determining potential categories according to interest categories; wherein the potential categories represent the content categories that are inferred to be of interest to the user to be recommended; and determining potential consumption content according to the potential categories.

[0079] Specifically, the interest category is the content category that the user to be recommended is interested in. After obtaining the interest category, the interest category can be expanded to obtain a potential category. The potential category is the content category that the user to be recommended may be interested in, that is, the content category that the user to be recommended is inferred to be interested in. For example, the semantic similarity between the interest category and all content categories can be calculated, and the N content categories with the highest similarity can be determined as potential categories. Alternatively, both the interest category and the N content categories with the highest similarity can be determined as potential categories.

[0080] From the preset historical content, multiple contents belonging to the potential category can be determined as potential consumption contents to be recommended to the user. For example, multiple potential consumption contents can be randomly selected. It is also possible to select historical contents with high popularity as potential consumption contents based on information such as the number of likes and views of the historical contents. For example, the higher the number of likes and / or views of the historical contents, the higher the popularity of the historical contents.

[0081] The beneficial effect of such a setting is that, based on the interest categories, potential categories that the recommended user may also be interested in are determined, thereby further achieving information gain, thereby being able to make personalized recommendations for users with sparse historical consumption information and improving user experience.

[0082] In this embodiment, determining the potential category according to the interest category includes: searching for the potential category associated with the interest category from a preset content category set.

[0083] Specifically, a content category set is preset, and the content category set may include all content categories. The content category set may also preset associations between different content categories, for example, there is an association between fitness tutorials and nutritional diets.

[0084] According to the preset association relationship, a content category associated with the interest category can be found from the content category set as a potential category. One interest category can be associated with one or more potential categories.

[0085] The beneficial effect of such a setting is that the association relationship between different content categories is pre-set, which facilitates the expansion of interest categories and realizes personalized recommendations for users.

[0086] In this embodiment, searching for potential categories associated with the interest category from a preset content category set includes: determining that the content category to which the content in the historical consumption information belongs is a known category; determining the similarity between the interest category and the known category; determining a target category from the interest category based on the similarity; wherein the similarity between the target category and the known category is greater than a preset similarity threshold; searching for potential categories associated with the target category from the preset content category set.

[0087] Specifically, the content consumed by the user to be recommended within a preset historical time period is determined as the consumed content. Each content corresponds to a content category, and the content category of the consumed content is determined as the known category.

[0088] An interest category is a content category related to the consumed content, and the interest category may include known categories or may not include known categories. For each interest category, the similarity between the interest category and the known category is determined, and the similarity can represent the degree of semantic similarity between the interest category and the known category. If there are multiple known categories, the similarity between an interest category and each known category can be calculated. In this embodiment, the method for calculating the similarity is not specifically limited.

[0089] A similarity threshold is preset, and the similarity corresponding to the interest category is compared with the preset similarity threshold. If one or more similarities corresponding to the interest category are greater than the preset similarity threshold, the interest category can be determined as the target category; if the similarities corresponding to the interest category are all less than the preset similarity threshold, the interest category is screened out, that is, the interest category is not the target category.

[0090] After determining the target category, the content category associated with the target category can be found from the preset content category set according to the association relationship between the preset content categories as the potential category. For example, if there are three interest categories A, B, and C, and the target category is only A, then only the content category associated with A needs to be found.

[0091] The beneficial effect of such a setting is that by calculating similarities, interest categories are sorted and screened, the amount of subsequent calculations is reduced, and the efficiency and accuracy of content recommendations are improved.

[0092] In this embodiment, potential consumption content is determined according to potential categories, including: determining the user group to which the recommended user belongs according to the user portrait of the recommended user; wherein the user group represents the set of all users corresponding to the user portrait; determining the recommended content corresponding to the user group; wherein the recommended content represents the content recommended to users in the user group in the preset historical content; and determining the potential consumption content from the recommended content according to the potential categories.

[0093] Specifically, the user portrait of the user to be recommended is determined. Different user portraits may correspond to different user groups. A user group may include multiple users, that is, a user group may represent a collection of all users corresponding to the user portrait. Based on the user portrait of the user to be recommended, the user group corresponding to the user portrait may be determined, that is, the user group to which the user to be recommended belongs may be determined.

[0094] For the user group to which the user to be recommended belongs, determine the recommended content corresponding to the user group. The recommended content represents the content recommended to any user in the user group in the preset historical content. That is, determine which content has been recommended to the users in the user group in the historical content, and determine the recommended content as the recommended content. Determine the content category corresponding to the recommended content, and determine the potential consumption content from the recommended content according to the potential category of the user to be recommended. For example, the content belonging to the potential category in the recommended content can be determined as the potential consumption content.

[0095] The beneficial effect of such a setting is to determine the content recommendations of other users who are similar to the user to be recommended, find out potential consumption content from the historically recommended content, improve the accuracy of determining potential consumption content, and thus improve the accuracy of content recommendation.

[0096] In this embodiment, potential consumption content is determined from recommended content according to potential categories, including: determining candidate consumption content from recommended content according to potential categories; wherein the content category of the candidate consumption content is the potential category; and determining a preset number of potential consumption content from the candidate consumption content.

[0097] Specifically, after determining the potential category, candidate consumption content is determined from the recommended content according to the potential category, and the content category of the candidate consumption content is the potential category. That is to say, from the recommended content, find the content of the potential category as the candidate consumption content. It is also possible to first determine the recommended content belonging to the potential category, and determine the consumption volume of the recommended content belonging to the potential category. The consumption volume can be data such as the number of views and the distribution volume. For example, the number of times the recommended content is distributed to all users is determined as the distribution volume. According to the consumption volume, the recommended content belonging to the potential category is sorted from high to low, and the content ranked in the top K positions is determined as the candidate consumption content.

[0098] From the candidate consumption contents, multiple contents are determined as potential consumption contents. For example, the number of potential consumption contents can be preset, and a preset number of contents can be randomly selected from the K candidate consumption contents as potential consumption contents. By performing random selection, it is possible to avoid returning the same content resources to users of similar user groups, thereby improving the diversity of recommendations.

[0099] The beneficial effect of such a setting is that candidate consumption content is first determined through potential categories, and then potential consumption content is determined from the candidate consumption content, thereby achieving double screening of the recommended content and improving the accuracy of content recommendation.

[0100] S204: Determine content to be recommended based on user information and potential consumption content.

[0101] Exemplarily, this step may refer to the above-mentioned step S103 and will not be described in detail.

[0102] S205: Push the content to be recommended to the user to be recommended.

[0103] Exemplarily, this step may refer to the above-mentioned step S104 and will not be described in detail.

[0104] In the disclosed embodiment, the user portrait and historical consumption information of the user to be recommended are obtained, and the content that the user may consume in the past large amount of historical content is inferred through the user portrait and known historical consumption information, as potential consumption content, to expand the user's performance. Then, based on the user portrait, as well as the known and inferred consumption content, the content that can be recommended to the user is determined. When the user's historical consumption information is insufficient, the content to be recommended is prevented from being affected by the popularity of the market, thereby realizing personalized recommendation of content, improving the accuracy of content recommendation, and enhancing user experience.

[0105] Figure 3 A flowchart of a content recommendation data processing method based on a large model provided in an embodiment of the present disclosure.

[0106] In this embodiment, content to be recommended is determined based on user information and potential consumption content, including: performing format conversion processing on user information and potential consumption content according to a preset data format to obtain data after format conversion processing; inputting the data after format conversion processing into a preset content recommendation model to obtain output content to be recommended; wherein the preset content recommendation model is used to perform feature extraction processing on input data in a preset data format, and output content to be recommended.

[0107] like Figure 3 As shown, the method comprises the following steps:

[0108] S301. Obtain user information of the user to be recommended; wherein the user information includes user portrait and historical consumption information within a historical time period.

[0109] Exemplarily, this step may refer to the above-mentioned step S101 and will not be described in detail.

[0110] S302: Determine potential consumption content according to user information and based on a preset macro model.

[0111] Exemplarily, this step may refer to the above-mentioned step S102 and will not be described in detail.

[0112] S303: Perform format conversion processing on the user information and the potential consumption content according to a preset data format to obtain format-converted data.

[0113] For example, after obtaining the potential consumption content, the content to be recommended can be determined by combining the user portrait, historical consumption information and potential consumption content. To ensure the real-time nature of content recommendation, the content to be recommended can be determined from the content published by the software product in the recent period of time. For example, the content that best matches the user information and the potential consumption content can be determined from the content uploaded and published to the software product by all users within 24 hours as the content to be recommended.

[0114] In order to improve the efficiency of data processing, the format of user information and potential consumption content can be unified. A data format is preset, and according to the preset data format, the user portrait, historical consumption information and potential consumption content are format converted respectively to obtain data after format conversion. That is, the user portrait, historical consumption information and potential consumption content after format conversion are obtained. For example, the preset data format can be a pb (protocol buffer) format.

[0115] S304: Input the format-converted data into a preset content recommendation model to obtain output content to be recommended; wherein the preset content recommendation model is used to perform feature extraction processing on the input data in a preset data format and output the content to be recommended.

[0116] Exemplarily, a content recommendation model is pre-built, and the content recommendation model can be a neural network model, the input data is user information and potential consumption content in a preset data format, and the output data is the content to be recommended. The content recommendation model can include multiple network layers, and can perform feature extraction and other processing on the input data. In this embodiment, the model structure of the content recommendation model is not specifically limited.

[0117] The format-converted data is input into the preset content recommendation model. The content recommendation model performs feature extraction on the input data, compares the extracted feature vector with the content in the database, and selects the content most similar to the feature vector as the content to be recommended. The database can store content recently uploaded or published by various users. For example, it can store content uploaded by various users within 24 hours. The content recommendation model can also be used to sort the content in the database, and the content ranked first can be used as the content to be recommended.

[0118] In this embodiment, format conversion is performed to facilitate data processing by the content recommendation model, thereby improving the accuracy and efficiency of data processing, realizing automatic screening and recommendation of content, improving the accuracy and efficiency of content recommendation, and enhancing user experience.

[0119] S305: Push the content to be recommended to the user to be recommended.

[0120] Exemplarily, this step may refer to the above-mentioned step S104 and will not be described in detail.

[0121] Figure 4 Schematic diagram of the content recommendation process. Figure 4 For new and light users, user portraits and historical consumption information can be obtained in real time or at a scheduled time. The historical consumption information can be a sequence of historical consumption content, and the elements in the sequence can be videos or pictures, etc. According to the user portrait and historical consumption sequence, personalized prompt words are obtained. According to the personalized prompt words and the preset big model, interest categories are obtained. According to the interest category, potential consumption content that the user may consume is obtained. Combined with the user portrait, historical consumption sequence, and potential consumption content, model inference is performed to obtain the content to be recommended. The model used in the model inference process is the content recommendation model.

[0122] In the disclosed embodiment, the user portrait and historical consumption information of the user to be recommended are obtained, and the content that the user may consume in the past large amount of historical content is inferred through the user portrait and known historical consumption information, as potential consumption content, to expand the user's performance. Then, based on the user portrait, as well as the known and inferred consumption content, the content that can be recommended to the user is determined. When the user's historical consumption information is insufficient, the content to be recommended is prevented from being affected by the popularity of the market, thereby realizing personalized recommendation of content, improving the accuracy of content recommendation, and enhancing user experience.

[0123] Figure 5 The structural block diagram of a content recommendation data processing device based on a large model provided by an embodiment of the present disclosure. For the convenience of explanation, only the parts related to the embodiment of the present disclosure are shown. Figure 5 The content recommendation data processing device 500 based on the large model includes: an information acquisition unit 501, a content determination unit 502, a recommendation determination unit 503 and a content push unit 504.

[0124] The information acquisition unit 501 is used to acquire user information of the user to be recommended; wherein the user information includes a user portrait and historical consumption information within a historical time period;

[0125] A content determination unit 502, configured to determine potential consumption content according to the user information and based on a preset macro model;

[0126] A recommendation determination unit 503, configured to determine content to be recommended according to the user information and the potential consumption content;

[0127] The content pushing unit 504 is used to push the content to be recommended to the user to be recommended.

[0128] Figure 6 A structural block diagram of a content recommendation data processing device based on a large model provided in an embodiment of the present disclosure, such as Figure 6 As shown, the content recommendation data processing device 600 based on the large model includes an information acquisition unit 601, a content determination unit 602, a recommendation determination unit 603 and a content push unit 604, wherein the content determination unit 602 includes a first determination module 6021 and a second determination module 6022.

[0129] The first determination module 6021 is used to determine the interest category according to the user information and based on a preset macro model; wherein the interest category is related to the category of historical consumption information;

[0130] The second determination module 6022 is used to determine potential consumption content according to the interest category.

[0131] In one example, the first determining module 6021 includes:

[0132] A prompt word generation submodule, used to generate prompt words according to the user information and based on a preset prompt word template;

[0133] The category determination submodule is used to input the prompt word into a preset large model to obtain an output interest category.

[0134] In one example, the category determination submodule is specifically used to:

[0135] Inputting the prompt word into the large model to obtain output semantic information of the prompt word;

[0136] An interest category is determined according to the semantic information.

[0137] In one example, the second determining module 6022 includes:

[0138] A potential determination submodule, used to determine a potential category according to the interest category; wherein the potential category represents the inferred content category that the user to be recommended is interested in;

[0139] The content determination submodule is used to determine the potential consumption content according to the potential category.

[0140] In one example, the potential determination submodule is specifically used for:

[0141] The potential category associated with the interest category is searched from a preset content category set.

[0142] In one example, the potential determination submodule is specifically used for:

[0143] Determining that the content category of the content in the historical consumption information is a known category;

[0144] determining a similarity between the interest category and the known category;

[0145] Determining a target category from the interest categories according to the similarity; wherein the similarity between the target category and the known category is greater than a preset similarity threshold;

[0146] The potential category associated with the target category is searched from a preset content category set.

[0147] In one example, the content determination submodule is specifically used to:

[0148] Determine the user group to which the user to be recommended belongs according to the user portrait of the user to be recommended; wherein the user group represents the set of all users corresponding to the user portrait;

[0149] Determine the recommended content corresponding to the user group; wherein the recommended content represents the content recommended to users in the user group in the preset historical content;

[0150] The potential consumption content is determined from the recommended content according to the potential category.

[0151] In one example, the content determination submodule is specifically used to:

[0152] Determining candidate consumption content from the recommended content according to the potential category; wherein the content category of the candidate consumption content is a potential category;

[0153] A preset number of potential consumption contents are determined from the candidate consumption contents.

[0154] In one example, the content push unit 603 includes:

[0155] A format conversion module, used to perform format conversion processing on the user information and the potential consumption content according to a preset data format to obtain format-converted data;

[0156] A content output module is used to input the data after the format conversion processing into a preset content recommendation model to obtain the output content to be recommended; wherein the preset content recommendation model is used to perform feature extraction processing on the input data in a preset data format and output the content to be recommended.

[0157] In one example, the user information includes historical consumption, where the historical consumption represents the amount of content consumed by the to-be-recommended user within a preset historical time period; the device further includes:

[0158] A first response unit is configured to, in response to the historical consumption being less than or equal to a preset quantity threshold, determine potential consumption content based on the user information and a preset macro model; determine content to be recommended based on the user information and the potential consumption content, and push the content to be recommended to the user to be recommended;

[0159] The second response unit is used to determine the content to be recommended according to the user information in response to the historical consumption being greater than a preset quantity threshold, and push the content to be recommended to the user to be recommended.

[0160] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device.

[0161] Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of the present disclosure. The electronic device may be a terminal device, a processor or a server. Figure 7As shown, the electronic device 70 includes: at least one processor 702; and a memory 701 communicatively connected to the at least one processor 702; wherein the memory stores instructions executable by the at least one processor 702, and the instructions are executed by the at least one processor 702 so that the at least one processor 702 can execute the large model-based content recommendation data processing method disclosed in the present invention.

[0162] The electronic device 70 further includes a receiver 703 and a transmitter 704. The receiver 703 is used to receive instructions and data sent by other devices, and the transmitter 704 is used to send instructions and data to external devices.

[0163] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0164] According to an embodiment of the present disclosure, the present disclosure also provides a computer program product, which includes: a computer program, the computer program is stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device executes the solution provided by any of the above embodiments.

[0165] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, 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 required herein.

[0166] like Figure 8 As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0167] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0168] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 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, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as a content recommendation data processing method based on a large model. For example, in some embodiments, the content recommendation data processing method based on a large model may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the content recommendation data processing method based on the large model described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to execute the large model-based content recommendation data processing method in any other appropriate manner (for example, by means of firmware).

[0169] Various implementations of the systems and techniques described above herein 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 chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations 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 that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0170] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0171] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0172] To provide 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 pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the 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 input, voice input, or tactile input).

[0173] The systems and techniques described herein may 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 with a graphical user interface or a web browser through which a user can interact with implementations 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 may 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.

[0174] A computer system may include a client and a server. The client and the server are generally remote 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 the corresponding 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 cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0175] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded 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 this document does not limit this.

[0176] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A content recommendation data processing method based on a large model, comprising: Obtaining user information of the user to be recommended; wherein the user information includes user portrait and historical consumption information within a historical time period; According to the user information, based on a preset macro model, determine potential consumption content; Determining content to be recommended based on the user information and the potential consumption content; The content to be recommended is pushed to the user to be recommended.

2. The method according to claim 1, wherein: The determining of potential consumption content according to the user information and based on a preset macro model includes: Determine an interest category according to the user information and based on a preset macro model; wherein the interest category is related to the category of the historical consumption information; The potential consumption content is determined according to the interest category.

3. The method according to claim 2, wherein: The determining of interest categories according to the user information and based on a preset macro model includes: Generate prompt words according to the user information and based on a preset prompt word template; The prompt word is input into a preset macro model to obtain the output interest category.

4. The method according to claim 3, wherein: The step of inputting the prompt word into a preset macro model to obtain the outputted interest category includes: Inputting the prompt word into the large model to obtain output semantic information of the prompt word; The interest category is determined according to the semantic information.

5. The method according to any one of claims 2 to 4, wherein: The determining the potential consumption content according to the interest category includes: Determine a potential category according to the interest category; wherein the potential category represents the inferred content category that the user to be recommended is interested in; The potential consumption content is determined according to the potential category.

6. The method according to claim 5, wherein: Determining potential categories according to the interest categories includes: The potential category associated with the interest category is searched from a preset content category set.

7. The method according to claim 6, wherein: The step of searching for the potential category associated with the interest category from a preset content category set includes: Determining that the content category of the content in the historical consumption information is a known category; determining a similarity between the interest category and the known category; Determining a target category from the interest categories according to the similarity; wherein the similarity between the target category and the known category is greater than a preset similarity threshold; The potential category associated with the target category is searched from a preset content category set.

8. The method according to any one of claims 5 to 7, wherein: The step of determining the potential consumption content according to the potential category includes: Determine the user group to which the user to be recommended belongs according to the user portrait of the user to be recommended; wherein the user group represents the set of all users corresponding to the user portrait; Determine the recommended content corresponding to the user group; wherein the recommended content represents the content recommended to users in the user group in the preset historical content; The potential consumption content is determined from the recommended content according to the potential category.

9. The method according to claim 8, wherein: The step of determining the potential consumption content from the recommended content according to the potential category includes: Determine candidate consumption content from the recommended content according to the potential category; wherein the content category of the candidate consumption content is a potential category; A preset number of potential consumption contents are determined from the candidate consumption contents.

10. The method according to any one of claims 1 to 9, wherein: The determining of the content to be recommended according to the user information and the potential consumption content includes: According to a preset data format, format conversion is performed on the user information and the potential consumption content to obtain format-converted data; The data after the format conversion processing is input into a preset content recommendation model to obtain the output content to be recommended; wherein the preset content recommendation model is used to perform feature extraction processing on the input data in a preset data format and output the content to be recommended.

11. The method according to any one of claims 1 to 10, wherein the user information includes historical consumption, and the historical consumption represents the amount of content consumed by the user to be recommended within a preset historical time period; The method further comprises: In response to the historical consumption amount being less than or equal to a preset quantity threshold, determining potential consumption content based on the user information and a preset macro model; determining content to be recommended based on the user information and the potential consumption content; and pushing the content to be recommended to the user to be recommended; In response to the historical consumption being greater than a preset quantity threshold, content to be recommended is determined according to the user information, and the content to be recommended is pushed to the user to be recommended.

12. A content recommendation data processing device based on a large model, comprising: An information acquisition unit, used to acquire user information of the user to be recommended; wherein the user information includes a user portrait and historical consumption information within a historical time period; A content determination unit, configured to determine potential consumption content according to the user information and based on a preset macro model; A recommendation determination unit, configured to determine content to be recommended based on the user information and the potential consumption content; The content push unit is used to push the content to be recommended to the user to be recommended.

13. The device according to claim 12, wherein: The content determination unit includes: A first determination module is used to determine an interest category according to the user information and based on a preset macro model; wherein the interest category is related to the category of the historical consumption information; The second determination module is used to determine the potential consumption content according to the interest category.

14. The device according to claim 13, wherein: The first determining module includes: A prompt word generation submodule, used to generate prompt words according to the user information and based on a preset prompt word template; The category determination submodule is used to input the prompt word into a preset macro model to obtain the output interest category.

15. The device according to claim 14, wherein: The category determination submodule is specifically used for: Inputting the prompt word into the large model to obtain output semantic information of the prompt word; The interest category is determined according to the semantic information.

16. The device according to any one of claims 13 to 15, wherein: The second determining module includes: A potential determination submodule, used to determine a potential category according to the interest category; wherein the potential category represents the inferred content category that the user to be recommended is interested in; The content determination submodule is used to determine the potential consumption content according to the potential category.

17. The device according to claim 16, wherein: The potential determination submodule is specifically used for: The potential category associated with the interest category is searched from a preset content category set.

18. The device according to claim 17, wherein: The potential determination submodule is specifically used for: Determining that the content category of the content in the historical consumption information is a known category; determining a similarity between the interest category and the known category; Determining a target category from the interest categories according to the similarity; wherein the similarity between the target category and the known category is greater than a preset similarity threshold; The potential category associated with the target category is searched from a preset content category set.

19. The device according to any one of claims 16 to 18, wherein: The content determination submodule is specifically used for: Determine the user group to which the user to be recommended belongs according to the user portrait of the user to be recommended; wherein the user group represents the set of all users corresponding to the user portrait; Determine the recommended content corresponding to the user group; wherein the recommended content represents the content recommended to users in the user group in the preset historical content; The potential consumption content is determined from the recommended content according to the potential category.

20. The device according to claim 19, wherein The content determination submodule is specifically used for: Determine candidate consumption content from the recommended content according to the potential category; wherein the content category of the candidate consumption content is a potential category; A preset number of potential consumption contents are determined from the candidate consumption contents.

21. The device according to any one of claims 12 to 20, wherein: The content pushing unit includes: A format conversion module, used to perform format conversion processing on the user information and the potential consumption content according to a preset data format to obtain format-converted data; A content output module is used to input the data after the format conversion processing into a preset content recommendation model to obtain the output content to be recommended; wherein the preset content recommendation model is used to perform feature extraction processing on the input data in a preset data format and output the content to be recommended.

22. The device according to any one of claims 12 to 21, wherein the user information includes historical consumption, and the historical consumption represents the amount of content consumed by the user to be recommended within a preset historical time period; The device also includes: A first response unit is configured to, in response to the historical consumption being less than or equal to a preset quantity threshold, execute the steps of determining potential consumption content according to the user information and based on a preset macro model; determining content to be recommended according to the user information and the potential consumption content; and pushing the content to be recommended to the user to be recommended; The second response unit is configured to determine the content to be recommended according to the user information in response to the historical consumption being greater than a preset quantity threshold, and push the content to be recommended to the user to be recommended.

23. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed 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 perform the method according to any one of claims 1 to 11.

24. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-11.

25. A computer program product, wherein: The invention comprises a computer program, which implements the steps of the method according to any one of claims 1 to 11 when being executed by a processor.

Images