Recommendation method and related products

Abstract

The invention discloses a recommendation method and related products. The method comprises the steps that at least one piece of first feature information, a language model, n reference materials and labels of the n reference materials of a target user are obtained, the target user is a registered user in a first internet platform, n is a positive integer, and the n reference materials and the labels of the n reference materials are all from the first internet platform; the at least one piece of first feature information comes from the outside of the first internet platform; based on the language model and the at least one piece of first feature information, determining a first tag from the tags of the n reference materials, the first tag being a tag of a material in which the target user is interested; based on the first label, m first materials interested by the target user are determined from the n reference materials, and m is a positive integer smaller than or equal to n.

Description

Technical Field

[0001] This application relates to the field of recommendation technology, and in particular to a recommendation method and related products. Background Technology

[0002] Current technology typically uses the feature information of materials interacting with the user to match with materials to be recommended, identifying a second material with a high degree of matching with the aforementioned feature information for recommendation. Since materials interacting with the user are usually of interest to the user, the probability of materials identified based on the feature information of interacted materials being of interest to the user is relatively high. However, using current technology to identify materials of user interest can easily lead to poor diversity in the identified materials. Summary of the Invention

[0003] This application provides a recommended method and related products, wherein the related products include a recommended device, an electronic device, a computer-readable storage medium, and a computer program product.

[0004] Firstly, a recommendation method is provided, which includes:

[0005] The system acquires at least one first feature information of the target user, a language model, n reference materials, and tags of the n reference materials. The target user is a registered user on a first Internet platform, where n is a positive integer. The n reference materials and the tags of the n reference materials are all from the first Internet platform, and the at least one first feature information is from outside the first Internet platform.

[0006] Based on the language model and the at least one first feature information, a first tag is determined from the tags of the n reference materials, wherein the first tag is a tag of the material that the target user is interested in;

[0007] Based on the first tag, m first materials that the target user is interested in are determined from the n reference materials, where m is a positive integer less than or equal to n.

[0008] In any embodiment of this application, the first feature information is used to indicate the popularity of the interactive object, the popularity of the interactive object is used to indicate the degree to which the interactive object is liked by the user, and the interactive object is an object that has interacted with the target user;

[0009] Determining the first label from the labels of the n reference materials based on the language model and the at least one first feature information includes:

[0010] A second feature is determined from the at least one first feature, wherein the popularity of the interactive object indicated by the second feature is less than a first threshold;

[0011] Using the language model, a second tag matching the interactive object corresponding to the second feature information is determined from the tags of the n reference materials;

[0012] The first label is determined based on the second label.

[0013] In any embodiment of this application, the interaction object includes locations visited by the target user, and / or applications installed by the target user;

[0014] When the interactive object includes a location visited by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have visited the location;

[0015] When the interactive object includes an application installed by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have installed the application.

[0016] In any embodiment of this application, determining m first materials of interest to the target user from the n reference materials based on the first tag includes:

[0017] If the number of first feature information corresponding to the first tag is greater than 1, and / or the number of times the material with the first tag has interacted with the user is greater than a second threshold, then based on the first tag, the m first materials are determined from the n reference materials. In any embodiment of this application, the method further includes:

[0018] Obtain the weights of the m first materials, where the weights of the m first materials represent the degree of interest a user has in the first materials, and the weights of the m first materials are obtained based on data from the first Internet platform;

[0019] A first sequence is determined based on the m first materials and their weights, wherein the order of the first materials in the first sequence is related to their weights.

[0020] Based on the first sequence, a target material is determined from the n reference materials, and the target material is the material to be recommended to the target user.

[0021] In any embodiment of this application, the order of the first material in the first sequence is negatively correlated with the weight of the first material, the correlation between the target material and the second material is greater than the correlation between the target material and the third material, the second material is the material in the first sequence in the first order, the third material is the material in the first sequence in the second order, and the first order is less than the second order.

[0022] In any embodiment of this application, the weight of the first material is related to at least one of the following: the number of likes the first material receives, the number of favorites the first material receives, the number of comments the first material receives, the number of shares the first material receives, and the number of followers the publisher of the first material receives.

[0023] In conjunction with any embodiment of this application, determining the first sequence based on the m first materials and their weights includes:

[0024] The first sequence is determined based on the m first materials, the weights of the m first materials, and the third feature information of the target user. The first sequence includes the m first materials and the third feature information, which comes from the first Internet platform.

[0025] In any embodiment of this application, determining the second tag matching the interactive object corresponding to the second feature information from the tags of the n reference materials includes:

[0026] Extract the semantics of the interactive object corresponding to the second feature information to obtain semantic information;

[0027] Determine the probability that the labels of the n reference materials match the semantic information to obtain at least one probability;

[0028] Based on the at least one probability, the second label is determined from the labels of the n reference materials, and the probability corresponding to the second label is greater than a third threshold.

[0029] In conjunction with any embodiment of this application, obtaining at least one first feature information of the target user includes:

[0030] If the number of materials that have interacted with the target user in the data from the first Internet platform is less than or equal to a reference threshold, the at least one first feature information is obtained.

[0031] In conjunction with any embodiment of this application, the method further includes:

[0032] If the number of materials that have interacted with the target user in the data from the first Internet platform is greater than the reference threshold, the m first materials are determined from the n reference materials based on the third feature information of the target user, wherein the third feature information comes from the first Internet platform.

[0033] Secondly, a recommendation device is provided, the recommendation device comprising:

[0034] The acquisition unit is used to acquire at least one first feature information of a target user, a language model, n reference materials, and tags of the n reference materials. The target user is a registered user on a first Internet platform, where n is a positive integer. The n reference materials and the tags of the n reference materials are all from the first Internet platform, and the at least one first feature information is from outside the first Internet platform.

[0035] The determining unit is configured to determine a first tag from the tags of the n reference materials based on the language model and the at least one first feature information, wherein the first tag is a tag of a material that the target user is interested in.

[0036] The determining unit is further configured to, based on the first tag, determine m first materials that the target user is interested in from the n reference materials, where m is a positive integer less than or equal to n.

[0037] In any embodiment of this application, the first feature information is used to indicate the popularity of the interactive object, the popularity of the interactive object is used to indicate the degree to which the interactive object is liked by the user, and the interactive object is an object that has interacted with the target user;

[0038] The aforementioned determining unit is specifically used for:

[0039] A second feature is determined from the at least one first feature, wherein the popularity of the interactive object indicated by the second feature is less than a first threshold;

[0040] Using the language model, a second tag matching the interactive object corresponding to the second feature information is determined from the tags of the n reference materials;

[0041] The first label is determined based on the second label.

[0042] In any embodiment of this application, the interaction object includes locations visited by the target user, and / or applications installed by the target user;

[0043] When the interactive object includes a location visited by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have visited the location;

[0044] When the interactive object includes an application installed by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have installed the application.

[0045] In conjunction with any embodiment of this application,

[0046] The aforementioned determining unit is specifically used for:

[0047] If the number of first feature information items corresponding to the first tag is greater than 1, and / or the number of times a material with the first tag has interacted with a user is greater than a second threshold, then based on the first tag, the m first materials are determined from the n reference materials. In conjunction with any embodiment of this application,

[0048] The aforementioned determining unit is specifically used for:

[0049] Obtain the weights of the m first materials, where the weights of the m first materials represent the degree of interest a user has in the first materials, and the weights of the m first materials are obtained based on data from the first Internet platform;

[0050] A first sequence is determined based on the m first materials and their weights, wherein the order of the first materials in the first sequence is related to their weights.

[0051] Based on the first sequence, a target material is determined from the n reference materials, and the target material is the material to be recommended to the target user.

[0052] In any embodiment of this application, the order of the first material in the first sequence is negatively correlated with the weight of the first material, the correlation between the target material and the second material is greater than the correlation between the target material and the third material, the second material is the material in the first sequence in the first order, the third material is the material in the first sequence in the second order, and the first order is less than the second order.

[0053] In conjunction with any embodiment of this application, the weight of the first material is related to at least one of the following: the number of likes the first material receives, the number of favorites the first material receives, the number of comments the first material receives, the number of shares the first material receives, and the number of followers the publisher of the first material receives.

[0054] In any embodiment of this application, the target database further includes third characteristic information of the target user;

[0055] The aforementioned determining unit is specifically used for:

[0056] The first sequence is determined based on the m first materials, the weights of the m first materials, and the third feature information of the target user. The first sequence includes the m first materials and the third feature information, which comes from the first Internet platform.

[0057] In conjunction with any embodiment of this application, the above-mentioned recommended apparatus further includes a semantic extraction unit, used for:

[0058] Extract the semantics of the interactive object corresponding to the second feature information to obtain semantic information;

[0059] The aforementioned recommended device further includes a computing unit, used for:

[0060] Determine the probability that the labels of the n reference materials match the semantic information to obtain at least one probability;

[0061] The aforementioned determining unit is specifically used for:

[0062] Based on the at least one probability, the second label is determined from the labels of the n reference materials, and the probability corresponding to the second label is greater than a third threshold.

[0063] In conjunction with any embodiment of this application, the above-mentioned acquisition unit is specifically used for:

[0064] If the number of materials that have interacted with the target user in the data from the first Internet platform is less than or equal to a reference threshold, the at least one first feature information is obtained.

[0065] In conjunction with any embodiment of this application, the above-mentioned determining unit is further configured to:

[0066] If the number of materials that have interacted with the target user in the data from the first Internet platform is greater than the reference threshold, the m first materials are determined from the n reference materials based on the third feature information of the target user, wherein the third feature information comes from the first Internet platform.

[0067] Thirdly, an electronic device is provided, comprising: a processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein, when the processor executes the computer instructions, the electronic device performs as described in the first aspect and any of its embodiments.

[0068] Fourthly, another electronic device is provided, comprising: a processor, a transmitting device, an input device, an output device, and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein, when the processor executes the computer instructions, the electronic device performs as described in the first aspect and any of its embodiments.

[0069] Fifthly, a computer-readable storage medium is provided, wherein a computer program is stored therein, the computer program including program instructions that, when executed by a processor, cause the processor to perform the first aspect and any of its embodiments described above.

[0070] In a sixth aspect, a computer program product is provided, the computer program product comprising a computer program or instructions that, when the computer program or instructions are executed on a computer, cause the computer to perform the first aspect described above and any of its embodiments.

[0071] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application.

[0072] In this embodiment, the recommendation device, after obtaining a language model, n reference materials, the tags corresponding to the n reference materials, and the first feature information of the target user, utilizes the semantic understanding capability of the language model and the first feature information to determine the first tags of materials of interest to the target user from the tags of the n reference materials. Based on these first tags, it then determines m first materials of interest to the target user from the n reference materials. The target user is a registered user on a first internet platform. The n reference materials and their tags originate from the first internet platform, while the first feature information originates from outside the first internet platform. Through these steps, the recommendation device determines the first materials of interest to the target user based on feature information from a target user outside the first internet platform, avoiding the problem of low diversity in the determined materials caused by determining materials of interest based on feature information from a target user within the first internet platform. This improves the diversity of the first materials determined based on the first feature information. Attached Figure Description

[0073] To more clearly illustrate the technical solutions in the embodiments of this application or the background art, the accompanying drawings used in the embodiments of this application or the background art will be described below.

[0074] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0075] Figure 1 A flowchart illustrating a recommended method provided in an embodiment of this application;

[0076] Figure 2 A flowchart illustrating the execution of a recommended method provided in an embodiment of this application;

[0077] Figure 3 This is a schematic diagram of the structure of a recommended device provided in an embodiment of this application;

[0078] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0079] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0080] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0081] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0082] The execution subject of this application embodiment is a recommendation device, which can be any electronic device capable of executing the technical solutions disclosed in the method embodiments of this application. Optionally, the recommendation device can be one of the following: a computer or a server.

[0083] It should be understood that the method embodiments of this application can also be implemented by a processor executing computer program code. The embodiments of this application are described below with reference to the accompanying drawings. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating a recommended method provided in an embodiment of this application.

[0084] 101. Obtain at least one primary feature information of the target user, a language model, n reference materials, and the tags of the n reference materials.

[0085] In this embodiment, the first feature information is the feature information of the target user, who is a registered user on the first internet platform. Since the first feature information originates outside the first internet platform, it differs from the feature information of the target user originating from the first internet platform. For example, the feature information of the target user originating from the first internet platform includes the target user's gender and age. The first feature information also includes information about applications installed by the target user.

[0086] Optionally, the first feature information originates from a target storage path on a second internet platform, which is different from the first internet platform. For example, the first feature information might be music listened to by the target user on the second internet platform. The recommendation device does not have permission to access the target storage path if it is not authorized to do so. For example, the target storage path might be the storage path for data in a database on the second internet platform, where the database includes data generated through the second internet platform.

[0087] Optionally, when acquiring the first feature information, the recommendation device needs to obtain access permission to the target storage path. For example, if the target storage path is the storage path of data in the database of the second internet platform, the recommendation device sends a request to the server of the second internet platform to access the target storage path. Upon receiving permission from the server to access the target storage path, the recommendation device acquires the first feature information from the database of the second internet platform by accessing the target storage path.

[0088] In one possible implementation, when the first characteristic information of the target user includes the target user's personal information, the rules for processing the first characteristic information are clearly communicated using a label / information, and authorization is obtained through pop-up messages or by the individual uploading their first characteristic information. The processing of the first characteristic information may include information such as the processor of the first characteristic information, the purpose of the processing, the processing method, and the type of first characteristic information being processed.

[0089] In this embodiment, the n reference materials originate from a first internet platform, where n is an integer greater than or equal to 1. Optionally, the reference materials include one or more of the following: text, video, image, audio, and product. For example, the reference materials include text. Another example is that the reference materials include images. Yet another example is that the reference materials include video, images, and text.

[0090] In this embodiment, the tags of the n reference materials come from a first internet platform. It should be understood that the tags of the n reference materials are a set of tags corresponding to the n reference materials, and each reference material corresponds to at least one tag. Different recommended materials may have the same or different tags. Therefore, the n reference materials have a tags, where a is an integer greater than or equal to 1.

[0091] For example, among the n reference materials, there are material y1 and material y2. The labels corresponding to material y1 include label x1 and label x2, and the labels corresponding to material y2 include label x2, label x3, and label x4. The labels of the n reference materials include label x1, label x2, label x3, and label x4.

[0092] Understandably, language models possess semantic understanding capabilities. By utilizing the semantic understanding capabilities of language models, the degree of semantic matching between different pieces of information can be determined, thereby identifying semantically matched information.

[0093] Optionally, the language model mentioned above is a large language model (LLM), where LLM refers to a deep learning model trained on a large amount of text data that can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, dialogue, text summarization, and code generation.

[0094] In one implementation of acquiring first feature information, a language model, n reference materials, and labels for the n reference materials, the recommendation device receives the first feature information, the language model, the n reference materials, and the labels for the n reference materials input by an input component. The input component includes at least one of the following: a keyboard, a mouse, a touchscreen, a touchpad, or an audio input device.

[0095] In another implementation of acquiring the first feature information, language model, n reference materials, and tags of the n reference materials, the recommendation device receives the first feature information sent by the terminal. The terminal includes at least one of the following: a mobile phone, a computer, a tablet computer, or a server.

[0096] Optionally, the first feature information includes one or more of the following: location-based services (LBS) location information (hereinafter referred to as geolocation information), application list, music list, etc., wherein the geolocation information includes information on locations visited by the target user, the application list includes information on applications installed by the target user, and the music list includes information on music listened to by the target user.

[0097] Optionally, the first feature information is a data management platform (DMP) profile, which includes information such as geographic location information, application list, and music list.

[0098] In one possible implementation scenario, the data from the first internet platform also includes the number of materials that have interacted with the target user. The recommendation device determines the number of materials that have interacted with the target user in the data from the first internet platform based on a reference threshold. Specifically, if the number of materials that have interacted with the target user in the data from the first internet platform is greater than the reference threshold, then the number of materials that have interacted with the target user in the data from the first internet platform is relatively large. The recommendation device can directly obtain the target user's third feature information and, based on the target user's third feature information, determine m first materials that the target user is interested in from n reference materials. The third feature information comes from the first internet platform and represents information generated by the target user on the first internet platform. Because the number of materials that have interacted with the target user is relatively large, the richness of the information generated by the target user on the first internet platform is high, and the accuracy of the first materials determined based on the richer third feature information is high.

[0099] In one possible implementation, the third characteristic information includes information generated when the target user registers on the first internet platform. When it is necessary to obtain the target user's third characteristic information, the personal information processing rules are clearly communicated using identifiers / information, and the individual's authorization is obtained through pop-up messages or by the individual uploading their personal information. Personal information processing may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0100] If the number of materials that have interacted with the target user in the data from the first internet platform is less than or equal to a reference threshold, it indicates that the number of materials that have interacted with the target user is relatively small. In this case, the feature information extracted from the materials that have interacted with the target user is limited. When determining materials that the target user is interested in based on limited feature information, the probability of the determined materials being materials that the target user is interested in is low. Therefore, the recommendation device needs to use first feature information from the target user outside the first internet platform to determine the materials that the target user is interested in. Since the first feature information comes from outside the first internet platform, it means that the first feature information is different from the information from the first internet platform. Based on the first feature information, m first materials that the target user is interested in are determined from n materials to be recommended. This avoids the low probability of the determined materials being materials that the target user is interested in due to the limited feature information from the first internet platform, thereby improving the accuracy of the determined materials being materials that the target user is interested in.

[0101] For example, the characteristic information of the target user from the first internet platform includes the target user's gender and age. The first characteristic information includes information about the applications the target user has installed. This first characteristic information differs from the characteristic information of the target user from the first internet platform.

[0102] For example, the characteristic information of the target user from the first Internet platform includes materials that the target user has interacted with. The first characteristic information includes locations visited by the target user.

[0103] 102. Based on the language model and at least one first feature information, determine the first label from the labels of n reference materials.

[0104] In this embodiment of the application, the first tag is a tag for materials that the target user is interested in, determined based on the first feature information.

[0105] Understandably, let 'a' be the number of tags corresponding to n reference materials. From these 'a' tags, determine the tags that match at least one first feature information to obtain the first tags. The number of first tags is 'b', where 'b' is a positive integer less than or equal to 'a'. The first tags determined by different first feature information can be the same or different. For example, at least one first feature information includes information A and information B. The first tags determined based on information A include tag x1 and tag x2. The first tags determined based on information B include tag x2 and tag x3. Therefore, the first tags determined based on at least one first feature information include tag x1, tag x2, and tag x3.

[0106] It should be understood that the number of first labels determined by different first feature information can be the same or different, and this application does not limit this. For example, at least one first feature information includes information A and information B. The first label determined based on information A includes label x1 and label x2. The first label determined based on information B includes label x4.

[0107] In this embodiment, the recommendation device utilizes the semantic understanding capability of a language model to determine the semantic matching degree between the first feature information and the tags corresponding to n reference materials, thereby determining the first tag based on the tag with the higher matching degree. Optionally, preset prompt words are used to guide the large language model to process the first feature information and the tags of the n reference materials, and output the first tag that matches the first feature information. For example, if the first feature information includes information about the locations visited by the target user, and the visited locations include location w1 and location w2, the preset prompt words include: If a user has visited location w1 and location w2, which 5 tags are they most likely to like? Five personalized results are generated for each location, maintaining diversity as much as possible. The output is in list form, and the output results must be tags from the a first tags. Under the guidance of the above preset prompt words, the large language model determines the 5 first tags corresponding to location w1 and the 5 first tags corresponding to location w2.

[0108] In one possible implementation, the semantics of the first feature information are extracted to obtain feature semantic information, and the probability of matching the labels corresponding to n reference materials with the feature semantic information is determined to obtain at least one feature probability. Based on the at least one feature probability, a first label is determined from the labels of the n reference materials, where the feature probability corresponding to the first label is greater than a matching threshold. For example, the first feature information includes "art museum," and the semantics of the first feature information are extracted to obtain at least one feature semantic information, which includes architectural space, cultural carrier, and art medium. The labels corresponding to the n reference materials include art exhibition, painting, sculpture, and sports. The probability of matching "art exhibition" with at least one feature semantic information is 92%. The probability of matching "painting" with at least one feature semantic information is 87%. The probability of matching "sculpture" with at least one feature semantic information is 81%. The probability of matching "sports" with at least one feature semantic information is 32%. When the matching threshold is 80%, the first label includes "art exhibition," "painting," and "sculpture."

[0109] In another possible implementation, the data from a first internet platform includes *d* first topics, each corresponding to at least one tag. The semantics of at least one first feature information are extracted to obtain at least one feature semantic information. The probability of each first topic matching at least one feature semantic information is determined from the *d* first topics, resulting in *d* third probabilities. From the *d* third probabilities, probabilities greater than a topic threshold are determined, resulting in *e* fourth probabilities. *e* second topics are determined based on the topics corresponding to the *e* fourth probabilities, and first tags are determined based on the tags corresponding to the *e* second topics. For example, the first feature information includes "art museum." Extracting the semantics of the first feature information yields at least one feature semantic information including "architectural space," "cultural carrier," and "art medium." The *d* first topics include "art exhibition," "painting," "sculpture," and "sports." The tags corresponding to "art exhibition" include "art," "culture," and "art exhibition." The tags corresponding to "painting" include "painting" and "photography." The tags corresponding to "sculpture" include "sculpture." The tags corresponding to "sports" include "football." The third probability of an art exhibition matching at least one feature semantic information is 92%. The third probability of a painting matching at least one feature semantic information is 87%. The third probability of a sculpture matching at least one feature semantic information is 81%. The third probability of a motion matching at least one feature semantic information is 32%. With a topic threshold of 80%, the fourth probabilities include 92%, 87%, and 81%. The e second topics corresponding to the fourth probability include art exhibitions, painting, and sculpture. The first labels determined based on the e second topics include art, culture, art exhibitions, painting, photography, and sculpture.

[0110] 103. Based on the first label, determine m first materials from n reference materials.

[0111] In this embodiment, the first material is a material that the target user is interested in. The recommendation device determines m first materials that the target user is interested in from n reference materials based on a first tag that represents the material. Here, m is a positive integer less than or equal to n. It should be understood that the first material includes the first tag. Since the first tag is a tag for materials that the target user is interested in, the probability that a first material including the first tag is a material that the target user is interested in is relatively high.

[0112] In this embodiment, the recommendation device, after obtaining a language model, n reference materials, the tags corresponding to the n reference materials, and the first feature information of the target user, utilizes the semantic understanding capability of the language model and the first feature information to determine the first tags of materials of interest to the target user from the tags of the n reference materials. Based on these first tags, it then determines m first materials of interest to the target user from the n reference materials. The target user is a registered user on a first internet platform. The n reference materials and their tags originate from the first internet platform, while the first feature information originates from outside the first internet platform. Through these steps, the recommendation device determines the first materials of interest to the target user based on feature information from a target user outside the first internet platform, avoiding the problem of low diversity in the determined materials caused by determining materials of interest based on feature information from a target user within the first internet platform, thereby improving the diversity of the first materials determined based on the first feature information.

[0113] Optionally, a more diverse first material can cover materials that the target user is interested in, thereby improving the accuracy of the first material in representing materials that the target user is interested in.

[0114] In one possible implementation scenario, the target user is a cold-start user, where a cold-start user is defined as a newly registered user on the first internet platform, or a user who has interacted with fewer items than a reference threshold since registering on the first internet platform. In this case, the probability that the recommendation device determines items of interest to the target user based on items they have interacted with is low. Therefore, the recommendation device obtains first feature information generated by the target user on platforms other than the first internet platform as a supplement to the user's feature information. The probability that the items determined based on this supplementary first feature information are items of interest to the target user is high, thus improving the accuracy of the recommendation device in determining items of interest to the target user when the target user is a cold-start user.

[0115] As an optional implementation, the first feature information is used to indicate the popularity of the interactive object, which is an object that has interacted with the target user.

[0116] Optionally, the popularity of an interactive object includes the number of times the interactive object is associated with a user. The fewer times an interactive object is associated with a user, the lower its popularity, and the higher the confidence that the interactive object is something the target user is interested in.

[0117] When the interactive object includes locations visited by the target user, the popularity of the interactive object is positively correlated with the number of users who have visited those locations. For example, the locations visited by the target user include museums and shopping malls. The number of users who visited shopping malls is 200,000, and the number of users who visited museums is 20,000. The greater number of users who visited shopping malls indicates higher popularity for shopping malls. Conversely, the smaller number of users who visited museums indicates lower popularity for museums.

[0118] When the interactive object includes an application that the target user has installed, the popularity of the interactive object is positively correlated with the number of users who have installed the aforementioned application.

[0119] When the interactive object includes music that the target user has listened to, the popularity of the interactive object is positively correlated with the number of users who have listened to the aforementioned music.

[0120] It should be understood that because the interactive object is something the target user has interacted with, the target user is highly likely to be interested in that interactive object. For example, if the interactive object includes a location the target user has visited, the target user is highly likely to be interested in that location. Similarly, if the interactive object includes an application the target user has installed, the target user is highly likely to be interested in that application. And if the interactive object includes music the target user has listened to, the target user is highly likely to be interested in that music.

[0121] It should be understood that different primary feature information corresponds to different interaction objects. For example, primary feature information includes feature information 1 and feature information 2. Feature information 1 indicates the popularity of interaction object 1. Feature information 2 indicates the popularity of interaction object 2. Interaction object 1 and interaction object 2 are different objects. For example, interaction object 1 is a museum visited by the target user, and interaction object 2 is a cinema visited by the target user.

[0122] The recommended device performs the following steps during step 102:

[0123] 201. Determine the second feature information from at least one first feature information.

[0124] The second feature information indicates that the popularity of the interactive object is less than the first threshold, where the popularity of the interactive object indicates the degree to which the interactive object is popular.

[0125] In one possible implementation, the interactive object is a location visited by the target user, and the popularity of the interactive object includes the number of users who have visited that location. If the number of users who have visited that location is less than a first threshold, the popularity of the interactive object is less than the first threshold.

[0126] In another possible implementation, the interactive object is an application that the target user has installed, and the popularity of the interactive object includes the number of users who have installed that application. If the number of users who have installed the application is less than a second threshold, the popularity of the interactive object is less than a first threshold.

[0127] In this embodiment, the lower the popularity of the interactive object indicated by the first feature information, the higher the probability that the interactive object is an unpopular object. Compared to popular objects, unpopular objects have a lower probability of being exposed. Therefore, if the interactive object is an unpopular object and the target user has interacted with it, the confidence level that the interactive object is an object of interest to the target user is high.

[0128] Therefore, in this embodiment of the application, the first threshold is used to filter the first feature information, and the popularity of the interactive object indicated by the second feature information is less than the first threshold. The confidence that the interactive object corresponding to the second feature information is an object of interest to the target user is relatively high.

[0129] 202. Using a language model, determine the second label that matches the interaction object corresponding to the second feature information from the labels of n reference materials.

[0130] In this embodiment, the recommendation device filters the first feature information using a first threshold to obtain second feature information, and uses the semantic understanding capability of a language model to determine a second tag matching the interaction object corresponding to the second feature information from the tags corresponding to n reference materials. The popularity of the interaction object indicated by the second feature information is lower than the first threshold. It should be understood that the number of tags corresponding to the n reference materials is greater than or equal to 1, and the number of second tags is less than or equal to the number of tags corresponding to the n reference materials.

[0131] Since the confidence level of the interaction object corresponding to the second feature information is that it is an object of interest to the target user is high, the probability that the second label determined by the interaction object corresponding to the second feature information is a label of material of interest to the target user is high.

[0132] Optionally, the semantics of the interaction object corresponding to the second feature information are extracted to obtain semantic information, and the probability of matching the tags of n reference materials with the semantic information is determined, resulting in at least one probability. Based on at least one probability, a second tag is determined from the tags of the n reference materials, where the probability corresponding to the second tag is greater than a third threshold. When the probability corresponding to the second tag is greater than the third threshold, the probability of the second tag matching the semantic information is relatively high. Since the semantic information is determined based on the interaction object corresponding to the second feature information, and the confidence that the interaction object corresponding to the second feature information is an object of interest to the target user is relatively high, the probability that the second tag determined based on the third threshold is a tag of a material of interest to the target user is relatively high.

[0133] 203. Determine the first label based on the second label.

[0134] In this embodiment of the application, the second tag is more likely to be a tag for materials that the target user is interested in, and the first tag determined based on the second tag is more likely to be a tag for materials that the target user is interested in.

[0135] In one possible implementation, the second tag is used as the first tag, where the first tag includes the tags in the second tag. For example, the second tag includes tags x1 and x2, and correspondingly, the first tag also includes tags x1 and x2.

[0136] In another possible implementation, tags that have been interacted with by the user more than a third threshold are identified from the second tags and used as the first tag. The third threshold is used to filter the number of interactions between the reference materials with the second tag and the user, thereby identifying reference materials with the second tag that have had a higher number of interactions with the user. The more interactions with the user, the higher the user's interest in the reference material, and the higher the probability that the target user is interested in the reference material. Through these steps, the probability that the identified first tag is a tag of material that the target user is interested in is further increased, thereby improving the accuracy of identifying the first material based on the first tag as material that the target user is interested in.

[0137] In this embodiment, the recommendation device filters first feature information based on a first threshold, determining feature information with lower popularity for indicated interactive objects from at least one set of first feature information. The lower the popularity of an interactive object, the higher the probability that it is an unpopular object. Compared to popular objects, unpopular objects have a lower probability of being exposed. Therefore, if the interactive object is an unpopular object and the target user has interacted with it, the confidence level of the interactive object being an object of interest to the target user is high. Based on this, the confidence level of the interactive object corresponding to the second feature information being an object of interest to the target user is high, the accuracy of the second tag determined based on the highly confident interactive object being a tag for materials of interest to the target user is high, the accuracy of the first tag determined based on the second tag being a tag for materials of interest to the target user is high, and the accuracy of the first material determined based on the highly accurate first tag being a material of interest to the target user is high.

[0138] As an optional implementation, it is recommended that the device perform the following steps during step 103:

[0139] 301. If the number of first feature information corresponding to the first tag is greater than 1, and / or the number of times the reference material with the first tag has interacted with the user is greater than a second threshold, then based on the second tag, determine m first materials from n reference materials.

[0140] In this embodiment, since the first label is a label determined based on at least one first feature information, the first label corresponds to at least one first feature information. For example, the at least one first feature information includes information A and information B. The first label determined based on information A includes label x1 and label x2. The first label determined based on information B includes label x2 and label x3. Label x1 corresponds to information A, and the target quantity of label x1 is 1. Label x2 corresponds to information A and information B, and the target quantity of label x2 is 2. Label x3 corresponds to information B, and the target quantity of label x3 is 1.

[0141] It should be understood that the higher the number of first feature information corresponding to the first tag, the higher the probability that the first tag is a tag of material that the target user is interested in.

[0142] Based on this, when the number of first feature information corresponding to the first tag is greater than 1, the probability that the first tag is a tag of material that the target user is interested in is relatively high, and the accuracy of determining m first materials as materials that the target user is interested in based on the second tag corresponding to the target number greater than 1 is relatively high.

[0143] Optionally, the first material includes at least one first label, and the number of first feature information corresponding to the first label is greater than 1. For example, m first materials include material y1 and material y2. Material y1 includes label x1 and label x2. Material y2 includes label x3. The first label includes label x1, label x2, and label x3, and the number of first feature information corresponding to the first label (label x1, label x2, and label x3) is greater than 1.

[0144] In this embodiment, a second threshold is used to filter the number of interactions, thereby determining the first tag corresponding to reference materials with a high number of interactions. The number of interactions refers to the number of times a material with the first tag has interacted with the user. A higher number of interactions indicates a higher probability that the user has interacted with the material with the first tag, thus indicating a higher probability that the user is interested in the material with the first tag. A higher probability that the user is interested in the material with the first tag also increases the probability that the first tag is a tag for a material the user is interested in, and further increases the probability that the first tag is a tag for a material the target user is interested in. Therefore, by determining the first tag of reference materials with a high number of interactions with the user based on the second threshold, and determining m first materials based on the first tag, the accuracy of identifying the first materials as materials the target user is interested in can be improved.

[0145] For example, tag x1 has 100,000 interactions, and tag x2 has 20,000 interactions. The probability that tag x1 is a tag for materials that the target user is interested in is higher than the probability that tag x2 is a tag for materials that the target user is interested in. With a second threshold of 40,000, based on tag x1, m first materials with tag x1 are determined from n reference materials, where tag x1 is the first tag.

[0146] In one possible implementation scenario, combining the above two methods can further improve the accuracy of determining that the first label of a first material is a material of interest to the target user. Specifically, when the number of first feature information corresponding to the first label is greater than 1, and the number of times the reference material with the first label has interacted with the user is greater than a second threshold, the number of first feature information corresponding to the first label and the target user is relatively large, and the number of times the material with the first label has interacted with the user is relatively large. Based on the judgment of the above two conditions, the probability of the first label being a label of a material of interest to the target user can be further increased, thereby further improving the accuracy of determining that the m first materials determined based on the first label are materials of interest to the target user.

[0147] In this embodiment, when the number of first feature information corresponding to the first tag is greater than 1, and / or the number of interactions between the user and the material with the first tag is greater than a second threshold, the recommendation device determines m first materials from n reference materials based on the first tag. The number of first feature information corresponding to the first tag and the number of interactions between the user and the material with the first tag both indicate the probability that the target user is interested in the material with the first tag. Specifically, the higher the number of first feature information corresponding to the first tag, the higher the probability that the target user is interested in the material with that first tag. Similarly, the higher the number of interactions between the user and the material with the first tag, the higher the probability that the target user is interested in the material with that first tag. Therefore, when the first tag satisfies the condition that the number of first feature information corresponding to the first tag is greater than 1, and / or the number of interactions between the user and the reference material with the first tag is greater than a second threshold, the probability that the first tag is a tag for materials of interest to the target user is high, and the accuracy of the m first materials determined based on the first tag as materials of interest to the target user is high.

[0148] As an optional implementation, the recommended device performs the following steps after performing step 103:

[0149] 401. Obtain the weights of m first materials.

[0150] In this embodiment, the weight of the first material represents the degree of user interest in the first material. Optionally, the higher the weight of the first material, the higher the degree of user interest in the first material.

[0151] Optionally, the weight of the first piece of content is related to at least one of the following: the number of likes, the number of favorites, the number of comments, the number of shares, and the number of followers of the publisher of the first piece of content. The formula for calculating the weight of the first piece of content is as follows:

[0152] ces=sum(like)+sum(fav)+sum(4*cmt)+sum(4*share)+sum(4*follow)…Formula (1)

[0153] Where, sum(like) is the number of likes the first piece of content received. sum(fav) is the number of favorites the first piece of content received. sum(4*cmt) is 4 times the number of comments the first piece of content received. sum(4*share) is 4 times the number of shares the first piece of content received. sum(4*follow) is 4 times the number of followers the publisher of the first piece of content received. ces is the weight of the first piece of content.

[0154] 402. Determine the first sequence based on the m first materials and their weights.

[0155] In this embodiment of the application, the first sequence includes m first materials, and the order of the first materials in the first sequence is related to the weight of the first materials.

[0156] Since the weight of the first item represents the degree of user interest in it, a higher weight for the first item indicates a higher level of user interest, and consequently, a higher level of user interest in the target user. The order of the first items in the first sequence is related to their weights; therefore, the order of the first items in the first sequence is related to the target user's level of interest in them. Differences in the order of different first items in the first sequence represent differences in the target user's level of interest in the different first items within the first sequence.

[0157] 403. Based on the first sequence, determine the target material to recommend to the target user from n reference materials.

[0158] In this embodiment, the target material is the material to be recommended to the target user. Since all m first materials in the first sequence are materials that the target user is interested in, the probability that the target material determined based on the m first materials in the first sequence is a material that the target user is interested in is relatively high. Recommending the target material that the target user is more likely to be interested in to the target user can improve the accuracy of the recommendation.

[0159] Optionally, the weight of any one of the m first materials is greater than the sixth threshold. Materials with a weight greater than the sixth threshold are those that the user is more interested in. Target materials determined based on materials that the user is more interested in have higher accuracy in identifying materials that the target user is interested in. Therefore, the weights of all m first materials determined by the recommendation device are greater than the sixth threshold, which avoids the presence of materials with low interest to the target user among the m first materials, thereby improving the accuracy of identifying target materials that the target user is interested in.

[0160] In this embodiment, the recommendation device obtains the weights of m first materials and determines a first sequence based on the m first materials and their weights. The order of the first materials in the first sequence is related to their weights, and each weight represents the degree of interest a user has in that first material. A higher weight indicates a higher degree of interest from the user, and consequently, a higher degree of interest from the target user. Since the order of the first materials in the first sequence is related to their weights, it is also related to the degree of interest the target user has in that first material. After determining the first sequence, the recommendation device selects target materials to recommend to the target user from n reference materials. During this process, the difference in the order of different first materials in the first sequence determines the difference in the target user's interest in different first materials, and thus identifies materials with a higher match to the target user, improving the accuracy of the target user's interest in the target materials.

[0161] In one possible implementation scenario, the order of the first material in the first sequence is negatively correlated with its weight; that is, the higher the weight of the first material, the lower its order in the first sequence, and the earlier it appears in the sequence. When generating the target material based on the first sequence, the target material has a higher relevance to the first material that appears earlier in the first sequence. Through the above steps, the influence of the later-ordered first material in the first sequence on the determination of the target material is reduced, thereby reducing the influence of the first material with a lower weight on the determination of the target material. Since the weight of the first material represents the degree of user interest in it, reducing the influence of the lower-weighted material on the determination of the target material can improve the accuracy of identifying the target material as something that the target user is interested in. For example, if the relevance between the target material and the second material is greater than the relevance between the target material and the third material, the second material is the first-ordered material in the first sequence, and the third material is the second-ordered material in the first sequence, with the first order being less than the second order. For example, if the first order is 2 and the second order is 3, the second material is the second material in the first sequence, and the third material is the third material in the first sequence.

[0162] Optionally, the first sequence also includes the third characteristic information of the target user. Including the third characteristic information and m first materials in the first sequence increases the richness of information in the first sequence, thereby improving the matching degree between the determined target materials and the target user, and increasing the accuracy of identifying the target materials as materials of interest to the target user. The third characteristic information of the target user originates from a first internet platform.

[0163] Optionally, the third feature information includes materials that have been interacted with by the target user.

[0164] Please see Figure 2 , Figure 2 The flowchart illustrates the execution of a recommended method provided in this application.

[0165] First, the recommendation device uses the language model data processing module to obtain at least one first feature information of the target user, and generates a preset prompt word based on the at least one feature information to guide the language model to match the first feature information with the first tag.

[0166] Specifically, the language model data processing module in the recommendation device acquires at least one first feature information of the target user, wherein the at least one first feature information includes the target user's geographical location information and an application list, wherein the geographical location information includes an art gallery. The application list includes application 1.

[0167] After receiving at least one feature, the recommendation device generates preset prompts based on that feature to guide the language model to match the first tag with the first feature. The preset prompts are as follows: If a user has visited an art gallery and installed application 1, which 5 tags are they most likely to like? Five personalized results are generated for each location and each application, maintaining as much diversity as possible. The output is in list format, and the output results are always tags from n reference materials.

[0168] Guided by preset prompts, the language model outputs five tags corresponding to the art museum: painting, art, art exhibition, culture, and photography. It also outputs five tags corresponding to application 1: photography, culture, video, entertainment, and lifestyle documentation. The first set of tags includes the five tags corresponding to the art museum and the five tags corresponding to application 1.

[0169] When generating a first tag corresponding to at least one first feature information, the user preference extraction module in the recommendation device is used to determine the number of first tags corresponding to the first feature information and the number of times that reference materials with the first tags have interacted with the user.

[0170] If the number of first tags corresponding to first feature information is greater than 1, the two first tags with the largest number of interactions between the reference material and the user are determined from the first tags with the largest number of interactions corresponding to first feature information, and these are used as user preference tags.

[0171] If the number of first labels corresponding to first feature information is equal to 1, determine the two second labels with the largest number of interactions from the first label as user preference labels.

[0172] Based on the steps, the recommendation device determines two first tags (culture and photography) as user preference tags from the five tags corresponding to the art museum (painting, art, art exhibition, culture, photography) and the five tags corresponding to the application 1 (photography, culture, video, entertainment, life record). The user preference tags include user preference tag 1 and user preference tag 2, where user preference tag 1 is culture and user preference tag 2 is photography.

[0173] Given the user preference tags, the material recall module in the recommendation device determines the m first materials that the user is interested in based on the two user preference tags mentioned above.

[0174] Specifically, the material recall module of the recommendation device uses the two user preference tags (i.e., user preference tag 1 and user preference tag 2) to determine m first materials from n reference materials using a recall method. The m first materials include material 1, material 2, material 3, and material 4.

[0175] Given m first materials, the new user cold start module in the recommendation device obtains a first sequence based on the m first materials, their weights, and the target user's third feature information. The third feature information includes materials from data on a first internet platform that have interacted with the target user. The order of the first materials in the first sequence is related to their weights. The first sequence is applied to the recall phase to determine the target materials to recommend to the target user. Then, the target materials are applied to the ranking phase to determine their recommendation order, and finally, the target materials are recommended to the target user.

[0176] Specifically, based on m first materials, the weights of the m first materials, and the third feature information of the target user, a first sequence is obtained, where the first sequence includes material 1, material 2, material 3, and material 4. The third feature information includes materials that have interacted with the target user, including material 0. The first sequence is applied to the recall stage to determine the target materials to be recommended to the target user. Then, the target materials are applied to the ranking stage to determine the recommendation order of the target materials, and finally, the target materials are recommended to the target user.

[0177] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, on the personal information processing device, the personal information processing rules are clearly indicated / informed, and authorization is obtained from the user through pop-up information or by asking the user to upload their personal information; wherein, personal information processing may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0178] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0179] The methods of the embodiments of this application have been described in detail above, and the apparatus of the embodiments of this application is provided below.

[0180] Please see Figure 3 , Figure 3 This is a schematic diagram of a recommendation device provided in an embodiment of this application. The recommendation device 1 includes: an acquisition unit 11 and a determination unit 12. Wherein:

[0181] The acquisition unit 11 is used to acquire at least one first feature information of the target user, a language model, n reference materials, and tags of the n reference materials. The target user is a registered user in the first Internet platform, where n is a positive integer. The n reference materials and the tags of the n reference materials are all from the first Internet platform, and the at least one first feature information is from outside the first Internet platform.

[0182] The determining unit 12 is used to determine a first tag from the tags of the n reference materials based on the language model and the at least one first feature information, wherein the first tag is a tag of a material that the target user is interested in.

[0183] The determining unit 12 is further configured to determine, based on the first tag, m first materials that the target user is interested in from the n reference materials, where m is a positive integer less than or equal to n.

[0184] In any embodiment of this application, the first feature information is used to indicate the popularity of the interactive object, the popularity of the interactive object is used to indicate the degree to which the interactive object is liked by the user, and the interactive object is an object that has interacted with the target user;

[0185] The aforementioned determining unit 12 is specifically used for:

[0186] A second feature is determined from the at least one first feature, wherein the popularity of the interactive object indicated by the second feature is less than a first threshold;

[0187] Using the language model, a second tag matching the interactive object corresponding to the second feature information is determined from the tags of the n reference materials;

[0188] The first label is determined based on the second label.

[0189] In any embodiment of this application, the interaction object includes locations visited by the target user, and / or applications installed by the target user;

[0190] When the interactive object includes a location visited by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have visited the location;

[0191] When the interactive object includes an application installed by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have installed the application.

[0192] In conjunction with any embodiment of this application, the determining unit 12 described above is specifically used for:

[0193] If the number of first feature information corresponding to the first tag is greater than 1, and / or the number of times a material with the first tag has interacted with a user is greater than a second threshold, then based on the first tag, the m first materials are determined from the n reference materials. In conjunction with any embodiment of this application, the determining unit 12 is specifically used for:

[0194] Obtain the weights of the m first materials, where the weights of the m first materials represent the degree of interest a user has in the first materials, and the weights of the m first materials are obtained based on data from the first Internet platform;

[0195] A first sequence is determined based on the m first materials and their weights, wherein the order of the first materials in the first sequence is related to their weights.

[0196] Based on the first sequence, a target material is determined from the n reference materials, and the target material is the material to be recommended to the target user.

[0197] In any embodiment of this application, the order of the first material in the first sequence is negatively correlated with the weight of the first material, the correlation between the target material and the second material is greater than the correlation between the target material and the third material, the second material is the material in the first sequence in the first order, the third material is the material in the first sequence in the second order, and the first order is less than the second order.

[0198] In conjunction with any embodiment of this application, the weight of the first material is related to at least one of the following: the number of likes the first material receives, the number of favorites the first material receives, the number of comments the first material receives, the number of shares the first material receives, and the number of followers the publisher of the first material receives.

[0199] In any embodiment of this application, the target database further includes third characteristic information of the target user;

[0200] The aforementioned determining unit 12 is specifically used to: determine the first sequence based on the m first materials, the weights of the m first materials, and the third feature information of the target user. The first sequence includes the m first materials and the third feature information, and the third feature information comes from the first Internet platform.

[0201] In conjunction with any embodiment of this application, the recommended device 1 further includes a semantic extraction unit 13, used for:

[0202] Extract the semantics of the interactive object corresponding to the second feature information to obtain semantic information;

[0203] The aforementioned recommended device 1 further includes a computing unit 14, used for:

[0204] Determine the probability that the labels of the n reference materials match the semantic information to obtain at least one probability;

[0205] The aforementioned determining unit 12 is specifically used for:

[0206] Based on the at least one probability, the second label is determined from the labels of the n reference materials, and the probability corresponding to the second label is greater than a third threshold.

[0207] In conjunction with any embodiment of this application, the above-mentioned acquisition unit 11 is specifically used for:

[0208] If the number of materials that have interacted with the target user in the data from the first Internet platform is less than or equal to a reference threshold, the at least one first feature information is obtained.

[0209] In conjunction with any embodiment of this application, the determining unit 12 is further configured to:

[0210] If the number of materials that have interacted with the target user in the data from the first Internet platform is greater than the reference threshold, the m first materials are determined from the n reference materials based on the third feature information of the target user, wherein the third feature information comes from the first Internet platform.

[0211] In this embodiment, the recommendation device, after obtaining a language model, n reference materials, the tags corresponding to the n reference materials, and the first feature information of the target user, utilizes the semantic understanding capability of the language model and the first feature information to determine the first tags of materials of interest to the target user from the tags of the n reference materials. Based on these first tags, it then determines m first materials of interest to the target user from the n reference materials. The target user is a registered user on a first internet platform. The n reference materials and their tags originate from the first internet platform, while the first feature information originates from outside the first internet platform. Through these steps, the recommendation device determines the first materials of interest to the target user based on feature information from a target user outside the first internet platform, avoiding the problem of low diversity in the determined materials caused by determining materials of interest based on feature information from a target user within the first internet platform, thereby improving the diversity of the first materials determined based on the first feature information.

[0212] In some embodiments, the functions or modules of the apparatus provided in this application can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0213] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. The electronic device 2 includes a processor 21 and a memory 22. Optionally, the electronic device 2 also includes an input device 23 and an output device 24. The processor 21, memory 22, input device 23, and output device 24 are coupled together via connectors, which include various interfaces, transmission lines, or buses, etc., and are not limited in this embodiment. It should be understood that in the various embodiments of this application, coupling refers to mutual connection in a specific way, including direct connection or indirect connection through other devices, such as through various interfaces, transmission lines, buses, etc.

[0214] Processor 21 may include one or more processors, such as one or more central processing units (CPUs). If the processor is a CPU, it may be a single-core CPU or a multi-core CPU. Optionally, processor 21 may be a processor group consisting of multiple CPUs, with the multiple processors coupled to each other via one or more buses. Optionally, the processor may also be other types of processors, etc., which are not limited in this embodiment.

[0215] The memory 22 can be used to store computer program instructions, as well as various types of computer program code, including program code for executing the scheme of this application. Optionally, the memory includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or compact disc read-only memory (CD-ROM), which is used for related instructions and data.

[0216] Input device 23 is used to input data and / or signals, and output device 24 is used to output data and / or signals. Input device 23 and output device 24 can be independent devices or an integrated device.

[0217] It is understood that in this embodiment of the application, the memory 22 can be used not only to store related instructions, but also to store related data. For example, the memory 22 can be used to store the first feature information obtained through the input device 23, or the memory 22 can also be used to store m first materials determined by the processor 21, etc. This embodiment of the application does not limit the specific data stored in the memory.

[0218] Understandable, Figure 4 This is merely a simplified design of an electronic device. In practical applications, the electronic device may also include other necessary components, including, but not limited to, any number of input / output devices, processors, memories, etc., and all electronic devices that can implement the embodiments of this application are within the protection scope of this application.

[0219] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0220] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. Those skilled in the art will also readily understand that the various embodiments of this application have different focuses, and for the sake of convenience and brevity, the same or similar parts may not be repeated in different embodiments. Therefore, parts not described or not described in detail in one embodiment can be referred to the descriptions in other embodiments.

[0221] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0222] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0223] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0224] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0225] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as read-only memory (ROM) or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A recommendation method, characterized in that, The method includes: The system acquires at least one first feature information of the target user, a language model, n reference materials, and tags of the n reference materials. The target user is a registered user on a first Internet platform, where n is a positive integer. The n reference materials and the tags of the n reference materials are all from the first Internet platform, and the at least one first feature information is from outside the first Internet platform. Based on the language model and the at least one first feature information, a first tag is determined from the tags of the n reference materials, wherein the first tag is a tag of the material that the target user is interested in; Based on the first tag, m first materials that the target user is interested in are determined from the n reference materials, where m is a positive integer less than or equal to n.

2. The method according to claim 1, characterized in that, The first feature information is used to indicate the popularity of the interactive object, and the popularity of the interactive object is used to indicate the degree to which the interactive object is liked by the user. The interactive object is an object that has interacted with the target user. Determining the first label from the labels of the n reference materials based on the language model and the at least one first feature information includes: A second feature is determined from the at least one first feature, wherein the popularity of the interactive object indicated by the second feature is less than a first threshold; Using the language model, a second tag matching the interactive object corresponding to the second feature information is determined from the tags of the n reference materials; The first label is determined based on the second label.

3. The method according to claim 2, characterized in that, The interaction objects include locations visited by the target user, and / or applications installed by the target user; When the interactive object includes a location visited by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have visited the location; When the interactive object includes an application installed by the target user, the popularity of the interactive object indicated by the first feature information is positively correlated with the number of users who have installed the application.

4. The method according to any one of claims 1-3, characterized in that, The step of determining m first materials of interest to the target user from the n reference materials based on the first tag includes: If the number of first feature information corresponding to the first tag is greater than 1, and / or the number of times the reference material with the first tag has interacted with the user is greater than a second threshold, the m first materials are determined from the n reference materials based on the first tag.

5. The method according to any one of claims 1-3, characterized in that, The method further includes: Obtain the weights of the m first materials, where the weights of the m first materials represent the degree of interest a user has in the first materials, and the weights of the m first materials are obtained based on data from the first Internet platform; A first sequence is determined based on the m first materials and their weights, wherein the order of the first materials in the first sequence is related to their weights. Based on the first sequence, a target material is determined from the n reference materials, and the target material is the material to be recommended to the target user.

6. The method according to claim 5, characterized in that, The order of the first material in the first sequence is negatively correlated with the weight of the first material. The correlation between the target material and the second material is greater than the correlation between the target material and the third material. The second material is the material in the first sequence in the first order, and the third material is the material in the second order in the first sequence. The first order is less than the second order.

7. The method according to claim 5, characterized in that, The step of determining the first sequence based on the m first materials and their weights includes: The first sequence is determined based on the m first materials, the weights of the m first materials, and the third feature information of the target user. The first sequence includes the m first materials and the third feature information, which comes from the first Internet platform.

8. The method according to claim 2 or 3, characterized in that, Determining the second tag from the tags of the n reference materials that matches the interactive object corresponding to the second feature information includes: Extract the semantics of the interactive object corresponding to the second feature information to obtain semantic information; Determine the probability that the labels of the n reference materials match the semantic information to obtain at least one probability; Based on the at least one probability, the second label is determined from the labels of the n reference materials, and the probability corresponding to the second label is greater than a third threshold.

9. The method according to any one of claims 1-3, characterized in that, The acquisition of at least one first feature information of the target user includes: If the number of materials that have interacted with the target user in the data from the first Internet platform is less than or equal to a reference threshold, the at least one first feature information is obtained.

10. The method according to claim 9, characterized in that, The method further includes: if the number of materials that have interacted with the target user in the data from the first Internet platform is greater than the reference threshold, determining the m first materials from the n reference materials based on the third feature information of the target user, wherein the third feature information comes from the first Internet platform.

11. A recommendation device, characterized in that, The recommendation device includes: The acquisition unit is used to acquire at least one first feature information of a target user, a language model, n reference materials, and tags of the n reference materials. The target user is a registered user on a first Internet platform, where n is a positive integer. The n reference materials and the tags of the n reference materials are all from the first Internet platform, and the at least one first feature information is from outside the first Internet platform. The determining unit is configured to determine a first tag from the tags of the n reference materials based on the language model and the at least one first feature information, wherein the first tag is a tag of a material that the target user is interested in; The determining unit is further configured to determine, based on the first tag, m first materials that the target user is interested in from the n reference materials, where m is a positive integer less than or equal to n.

12. An electronic device, characterized in that, include: A processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein, when the processor executes the computer instructions, the electronic device performs the method as described in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 10.

14. A computer program product, characterized in that, The computer program product includes a computer program or instructions; when the computer program or instructions are executed on a computer, the computer causes the computer to perform the method according to any one of claims 1 to 10.

Images