Information flow recommendation method and device, computer device and storage medium

Abstract

The application relates to an information flow recommendation method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining target attribute information corresponding to a target identifier, the target attribute information comprising an identifier category of the target identifier, the identifier category comprising a member and a non-member; determining a corresponding target recommendation label according to the identifier category of the target identifier, the target recommendation label being a member recommendation label or a non-member recommendation label, the member recommendation label being used for guiding a member to contribute video viewing time, and the non-member recommendation label being used for guiding a non-member to contribute member recharge conversion to a member; determining a to-be-recommended information flow based on the association degree between each candidate information flow and the target recommendation label; and pushing the to-be-recommended information flow to a terminal corresponding to the target identifier, so that different information flow pushing strategies are adopted for the target identifier of the member and the non-member to achieve different pushing targets.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an information flow recommendation method, apparatus, computer device, and storage medium. Background Technology

[0002] As users pay increasing attention to entertainment content, apps that provide such content have emerged. Users can subscribe to different membership levels within these apps to access content with varying access permissions. However, current apps with membership subscriptions use a uniform recommendation standard for their feeds, regardless of membership type, failing to consider the differences between membership types. The purpose of these feeds differs for different membership types. For example, members may have access to all videos and be expected to contribute more viewing time, while non-members may have access to only a few videos and be expected to subscribe to a membership to generate revenue. Using a uniform recommendation standard for different membership types makes it difficult to achieve the desired feed objectives for each membership type. Summary of the Invention

[0003] To address the aforementioned technical problems, this application provides an information flow recommendation method, apparatus, computer device, and storage medium.

[0004] Firstly, this application provides an information flow recommendation method, including:

[0005] Obtain target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members;

[0006] The corresponding target recommendation tag is determined according to the identification category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute member recharge to convert to member status.

[0007] The information stream to be recommended is determined based on the correlation between each candidate information stream and the target recommendation tag;

[0008] The recommended information stream is pushed to the terminal corresponding to the target identifier.

[0009] Secondly, this application provides an information flow recommendation device, comprising:

[0010] The acquisition module is used to acquire target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members;

[0011] The target determination module is used to determine the corresponding target recommendation tag according to the identifier category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute to the conversion of members to members through recharge.

[0012] The information flow determination module is used to determine the information flow to be recommended based on the correlation between each candidate information flow and the target recommendation tag;

[0013] The push module is used to push the information stream to be recommended to the terminal corresponding to the target identifier.

[0014] Thirdly, this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0015] Obtain target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members;

[0016] The corresponding target recommendation tag is determined according to the identification category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute member recharge to convert to member status.

[0017] The information stream to be recommended is determined based on the correlation between each candidate information stream and the target recommendation tag;

[0018] The recommended information stream is pushed to the terminal corresponding to the target identifier.

[0019] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0020] Obtain target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members;

[0021] The corresponding target recommendation tag is determined according to the identification category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute member recharge to convert to member status.

[0022] The information stream to be recommended is determined based on the correlation between each candidate information stream and the target recommendation tag;

[0023] The recommended information stream is pushed to the terminal corresponding to the target identifier.

[0024] Based on the above information flow recommendation method, target attribute information corresponding to the target identifier is obtained. The target attribute information includes the identifier category of the target identifier, which includes members and non-members. A corresponding target recommendation tag is determined according to the identifier category of the target identifier. The target recommendation tag is either a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute to member recharge conversion. Different information flow push strategies are adopted for member and non-member target identifiers to achieve different push objectives. Based on the correlation between each candidate information flow and the target recommendation tag, the information flow to be recommended is determined. According to the correlation between different candidate information flows and the target recommendation tags corresponding to different identifier categories, candidate information with higher correlation is determined as the information to be recommended, and the information flow to be recommended is pushed to the corresponding terminal of the target identifier. This solves the problem that the information flow push purpose corresponding to different member types cannot be achieved due to the use of unified tags and unified feature weight modeling for information flow push of different member types. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a diagram illustrating the application environment of an information flow recommendation method in one embodiment.

[0028] Figure 2 This is a flowchart illustrating an information flow recommendation method in one embodiment;

[0029] Figure 3This is a structural block diagram of an information flow recommendation device in one embodiment;

[0030] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] Figure 1 This is a diagram illustrating the application environment of an information flow recommendation method in one embodiment. (Refer to...) Figure 1 This information flow recommendation method is applied to an information flow recommendation system. The information flow recommendation system includes a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 has information flow browsing software installed. Users can log in to the information flow browsing software using a user ID to view information flow content. The information flow browsing software can be social media software, audio / video playback software, etc. In this embodiment, the information flow browsing software is video playback software. The data types of the information flow include text, images, audio, and video. The terminal 110 can be a desktop terminal or a mobile terminal. The mobile terminal can be at least one of a mobile phone, tablet computer, or laptop computer. The server 120 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0033] In one embodiment, Figure 2 This is a flowchart illustrating an information flow recommendation method in one embodiment, with reference to... Figure 2 This paper provides an information flow recommendation method. This embodiment mainly applies this method to the above-mentioned... Figure 1 Taking server 120 as an example, this information flow recommendation method specifically includes the following steps:

[0034] Step S210: Obtain target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members.

[0035] Specifically, the target identifier is any user identifier used to log in to the information flow browsing software. The user identifier can be a phone number, email address, third-party application account, or custom characters. Identity information includes [list of information types]. Custom characters can be any combination of numbers, letters, symbols, etc. Different user identifiers are used to indicate different users, and different users have different preferences for different information flow content. The target information is the attribute information corresponding to the target identifier; that is, each user identifier corresponds to one attribute information. Attribute information includes material information and identity information. Material information includes information flow ID, page views, browsing channel, information type, and other material characteristics. Identity information includes identifier category, gender, age, phone number, and other identity characteristics. Identifier category includes members and non-members. In this embodiment, the information flow browsing software is software with a membership subscription function. The information flow browsing software provides information flow content with different browsing permissions for users of different identifier categories. Browsing permissions include authorized browsing and unauthorized browsing. That is, different information flows have different browsing permissions for users of different identifier categories. For example, the same information flow may have authorized browsing permissions for members and unauthorized browsing permissions for non-members.

[0036] Step S220: Determine the corresponding target recommendation tag according to the identifier category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute to the conversion of members to members through recharge.

[0037] Specifically, different identifier categories correspond to different target recommendation tags. Target recommendation tags include member recommendation tags and non-member recommendation tags. Member recommendation tags or non-member recommendation tags include multiple behavioral tags. That is, member recommendation tags and non-member recommendation tags are used to indicate different push targets. The push target for members is to expect members to contribute more video viewing time; the push target for non-members is to convert the identifier category from non-member to member in order to contribute member recharge revenue.

[0038] Step S230: Determine the information stream to be recommended based on the correlation between each candidate information stream and the target recommendation tag.

[0039] Specifically, candidate information streams are those in the information database that are currently unviewed relative to the target identifier. To achieve the information stream push targets corresponding to different identifier categories, the correlation between each candidate information stream and the target recommendation tag is determined. A higher correlation indicates a greater contribution of the candidate information stream to achieving the corresponding push target after being pushed to the target identifier's corresponding terminal; conversely, a lower correlation indicates a lower contribution. Candidate information streams with high correlation are selected as the information streams to be recommended; that is, the information streams to be recommended include at least one candidate information stream.

[0040] Step S240: Push the information stream to be recommended to the terminal corresponding to the target identifier.

[0041] Specifically, the recommended information stream is pushed to the terminal corresponding to the target identifier to drive the behavior of the corresponding user to the corresponding push target. Different information stream push strategies are adopted for members and non-members to achieve different push goals, namely, to encourage non-members to recharge their membership fees to become members and contribute to member recharge revenue, or to encourage members to contribute more browsing time.

[0042] In one embodiment, determining the corresponding target recommendation tag based on the identifier category of the target identifier includes:

[0043] When the target identifier's identifier category is "member," a "member recommendation" tag is used as the target recommendation tag. This member recommendation tag includes a "click to play" tag, a first video viewing duration tag, and a second video viewing duration tag. The first video viewing duration tag indicates that the viewing duration of the first target video exceeds a preset percentage of the total video duration. The second video viewing duration tag indicates that the viewing duration of the second target video equals the total video duration. The first target video and the second target video can be any two different videos; or...

[0044] When the target identifier's identifier category is non-member, the non-member recommendation tag is used as the target recommendation tag. The non-member recommendation tag includes a click-to-play tag, a third video viewing duration tag, and an order page click tag. The third video viewing duration tag is used to indicate that the viewing duration of the third target video has reached a preset duration. The order page click tag is used to indicate that a non-member clicks on the member order page to contribute to the member recharge and convert to a member. The third target video and the first target video and the second target video can be any video that is the same or different.

[0045] Specifically, for users identified as members, a membership recommendation tag is created by combining three behavioral tags: a "click to play" tag, a "first video viewing time" tag, and a "second video viewing time" tag. These behavioral tags are categorized and labeled based on the browsing behavior of each user, meaning each information stream has at least one behavioral tag. The "click to play" tag indicates that the user clicks to play a video. The "first video viewing time" tag indicates that the user's viewing time for the first target video exceeds a preset percentage of the total video duration. This preset percentage can be configured based on the viewing time of member-level videos. For example, if the preset percentage is 0.75%, then the "first video viewing time" tag indicates that the user has watched more than 0.75% of the total video duration, meaning the user has watched most of the content in the first target video. The "second video viewing time" tag indicates that the user has watched the second target video completely. By setting these three behavioral tags, user behavior is driven towards these three tags, i.e., users are guided to contribute more video viewing time.

[0046] For users categorized as non-members, a non-member recommendation tag is created by combining three behavioral tags: a play button, a third-party video viewing duration tag, and an order page click tag. The third-party video viewing duration tag indicates that the viewing duration of the target video has reached a preset time, which can be 5 minutes, 6 minutes, 10 minutes, etc. In this embodiment, the preset time is set to 6 minutes. The order page click tag indicates that the user clicks on the member order page, thereby driving the user to convert their non-member status to member status and contribute to member recharge revenue.

[0047] In one embodiment, the target recommendation tag includes multiple behavioral tags, such as a click-to-play tag, a first video viewing duration tag, a second video viewing duration tag, a third video viewing duration tag, or an order page click tag. The step of determining the information stream to be recommended based on the correlation between each candidate information stream and the target recommendation tag includes:

[0048] The recommendation value corresponding to the candidate information stream is obtained by weighted summing of the correlation between the candidate information stream and each of the behavior tags.

[0049] The candidate information streams whose recommendation values ​​meet the recommendation criteria are taken as the information streams to be recommended, wherein the recommendation criteria include a recommendation value greater than or equal to a preset value, and / or, the recommendation value is located before or after a preset rank in the recommendation value ranking.

[0050] Specifically, the correlation between each candidate information stream and each behavior tag is determined. The correlation is used to indicate the contribution of the candidate information stream to the realization of the behavior tag. That is, the higher the correlation, the higher the contribution of the candidate information stream to the realization of the behavior tag. Pushing a candidate information stream with a higher correlation is more likely to achieve the push target of the corresponding tag category.

[0051] However, since the target recommendation tags include multiple behavioral tags, in order to obtain a unique recommendation value for each candidate information stream, the correlation between the candidate information stream and each behavioral tag is weighted and summed. The sum of the weighting coefficients for each behavioral tag is 1. The weighting coefficients for each behavioral tag can be customized according to actual business needs. This summarizes the multiple correlations between the candidate information stream and each behavioral tag to obtain the recommendation value corresponding to the candidate information stream. The recommendation values ​​of each candidate information stream are then sorted in ascending or descending order to obtain a recommendation value ranking. In ascending order, candidate information streams whose recommendation values ​​are after a preset ranking are selected as the information streams to be recommended; in descending order, candidate information streams whose recommendation values ​​are before a preset ranking are selected as the information streams to be recommended. Alternatively, candidate information streams whose recommendation values ​​are greater than or equal to a preset value are selected as the information streams to be recommended.

[0052] For example, for members, long videos that they are interested in and that have a relatively long playback time can be selected as recommended feeds to encourage users to contribute to the video viewing time; for non-members, videos that they are interested in and that require member playback permissions can be selected as recommended feeds to encourage users to switch from non-member subscriptions to member viewing to encourage users to contribute to member subscription revenue.

[0053] Based on different identifier categories, and according to different push objectives, combined with the interests and preferences corresponding to each user identifier, candidate information streams are categorized and sorted. This allows for the delivery of information stream content that satisfies user interests and achieves the push objectives to the terminals of users corresponding to different identifier categories. Specifically, different behavioral tags are defined for members and non-members, driving both groups towards their respective target paths. For non-members with limited playback, clicking on the member order page is the ultimate goal. A multi-objective path learning model reinforces this group's ability to click more often to enter the member order page, encouraging non-members to recharge and convert into members, thus contributing to member recharge revenue. For members with unlimited playback, watching the entire video is the ultimate goal. A multi-objective path learning model encourages members to watch more videos, contributing to greater video viewing time.

[0054] In one embodiment, before determining the information stream to be recommended based on the correlation between each candidate information stream and the target recommendation tag, the method further includes:

[0055] Based on the attribute description vector corresponding to the target attribute information, a corresponding feature weight vector is constructed, wherein the feature weight vector is used to indicate the preference of the target identifier for each attribute feature;

[0056] According to the target recommendation label, the embedding mapping results of the corresponding information description vectors and feature weight vectors of each candidate information stream are classified to obtain the correlation degree between each candidate information stream and the target recommendation label.

[0057] Specifically, the attribute description vector includes feature vectors corresponding to multiple attribute features. The attribute features include the aforementioned material features and / or identity features. The feature weight vector includes the target identifier's preference for each attribute feature. Based on the Attention mechanism, the attribute description vector is transformed into a feature weight vector. That is, based on the target attribute information analysis, the preference of the target identifier's corresponding user for different attribute features is determined. Different users have different preferences, so different users have different preferences for different attribute features. Different user identifiers correspond to different feature weight vectors, and the feature weight vector is denoted as the Embedding vector.

[0058] The embedding mapping process is performed on the information description vectors and feature weight vectors of each candidate information stream. Specifically, the feature weight vectors are used as the embedding network layer in the neural network. The information description vectors of the candidate information streams are input into the embedding network layer for embedding mapping, which weights the candidate information streams according to the browsing preferences of the target user, resulting in a weighted embedding mapping result, which is the embedding mapping vector. Then, the embedding mapping vectors corresponding to each candidate information stream are classified according to the target recommendation tag. Specifically, the embedding mapping vectors output from the embedding network layer are input into a classification network layer for classification. The classification network layer can employ classification model structures such as FM, DCA, DECN, or MLP cross-classification. The classification network layer outputs the correlation between the embedding mapping vector and the target recommendation tag.

[0059] In other words, during the training process, the aforementioned neural network is trained using sample data from the training set according to the aforementioned method of pushing information streams to members or non-members, until the training loss function reaches its minimum value and the accuracy reaches the preset accuracy. This results in a neural network model trained by deep learning, which is used to implement the information stream recommendation methods in the above embodiments. The sample data is data labeled with at least one behavioral tag. The neural network model can filter and sort information streams that achieve different push targets for member groups or non-member groups.

[0060] In one embodiment, constructing a corresponding feature weight vector based on the attribute description vector corresponding to the target attribute information includes:

[0061] The target attribute information is mapped and transformed to the corresponding attribute description vector, wherein the attribute description vector is composed of feature vectors corresponding to multiple attribute features;

[0062] Obtain preset key features, wherein the preset key features include at least one attribute feature among the plurality of attribute features;

[0063] A first weight vector is formed based on the feature vectors corresponding to the preset key features in the attribute description vector;

[0064] The feature weight vector is generated by combining the attribute description vector with the first weight vector.

[0065] Specifically, each attribute feature in the target attribute information is represented by a vector, resulting in feature vectors corresponding to multiple attribute features. The preset key features are pre-configured and can be customized according to actual business needs. The preset key features include at least one attribute feature. For example, the preset key features include video type and video upload time. Video type includes long video, short video, and medium video. Video upload time is used to distinguish whether a video is a newly uploaded video. Usually, highly discriminative attribute features are selected as preset key features. The feature vectors corresponding to the preset key features in the attribute description vector are combined to form a first weight vector. The first weight vector is the vector corresponding to the preset key features. The feature weight vector is generated by combining the attribute description vector and the first weight vector. That is, the feature vectors of other attribute features are weighted by the importance of the highly discriminative attribute features to obtain the weighted feature weight vector.

[0066] In one embodiment, generating the feature weight vector by combining the attribute description vector with the first weight vector includes:

[0067] The first weight vector and the attribute description vector are subjected to basic operations to obtain the second weight vector, wherein the basic operations include multiplication, subtraction and addition;

[0068] The combined vector formed by concatenating the second weight vector and the attribute description vector is used as the feature weight vector.

[0069] Specifically, the first weight vector is denoted as Query, and the attribute description vector is denoted as Key. The second weight is obtained by multiplying, adding, or subtracting the first weight vector from the attribute description vector. For different basic operations, the second weight vector includes a first sub-vector, a second sub-vector, and a third sub-vector. The first sub-vector indicates the result of multiplying the first weight vector from the attribute description vector, i.e., Query*Key. The second sub-vector indicates the result of adding the first weight vector from the attribute description vector, i.e., Query+Key. The third sub-vector indicates the result of subtracting the first weight vector from the attribute description vector, i.e., Query-Key.

[0070] The combined vector formed by concatenating the second weight vector and the attribute description vector can be used as the feature weight vector. Specifically, the result of multiplying the first sub-vector and the attribute description vector can be used as the feature weight vector, or the result of multiplying the second sub-vector and the attribute description vector can be used as the feature weight vector, or the result of multiplying the first sub-vector, the second sub-vector, the first weight vector, and the attribute description vector into a compressed vector and the result of multiplying it with the attribute description vector can be used as the feature weight vector. Other methods of forming the feature weight vector based on the combination of the second weight vector and the attribute description vector will not be elaborated here.

[0071] For example, if the attribute description vector includes feature vectors corresponding to 10 attribute features, and each feature vector has a dimension of 16, then the element dimension of the attribute description vector is 10*16. If the preset key features include 2 attribute features, concatenating the feature vectors corresponding to the 2 attribute features yields a vector with an element dimension of 2*16. This vector is then compressed to obtain a first weight vector with an element dimension of 1*16, meaning the element dimension of the first weight vector is 1*M, where M is the number of dimensions contained in each feature vector. Multiplying the first weight vector by the attribute description vector yields a first sub-vector with an element dimension of 1*10. Multiplying the first sub-vector by the attribute description vector yields a feature weight vector with an element dimension of 10*16.

[0072] In one embodiment, constructing a corresponding feature weight vector based on the attribute description vector corresponding to the target attribute information includes:

[0073] Based on the material description vector and / or identity description vector corresponding to the target attribute information, a corresponding feature weight vector is constructed, wherein the attribute description vector includes the material description vector and / or identity description vector.

[0074] Specifically, since the target attribute information includes material information and identity information, when constructing the feature weight vector, the feature weight vector can be constructed using only the material description vector corresponding to the material information, or only the identity description vector corresponding to the identity information, or the feature weight vector can be constructed by combining the material description vector and the identity description vector. That is, in the above embodiments, the attribute description vector is either the material description vector or the identity description vector, or it includes both the material description vector and the identity description vector.

[0075] Figure 2 This is a flowchart illustrating an information flow recommendation method in one embodiment. It should be understood that, although... Figure 2 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 2 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0076] In one embodiment, such as Figure 3 As shown, an information flow recommendation device is provided, comprising:

[0077] The acquisition module 310 is used to acquire target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members;

[0078] The target determination module 320 is used to determine the corresponding target recommendation tag according to the identification category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag, the member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute member recharge to convert to member status;

[0079] The information flow determination module 330 is used to determine the information flow to be recommended based on the correlation between each candidate information flow and the target recommendation tag;

[0080] The push module 340 is used to push the information stream to be recommended to the terminal corresponding to the target identifier.

[0081] In one embodiment, the target determination module 320 is specifically used for:

[0082] When the target identifier's identifier category is "member," a "member recommendation" tag is used as the target recommendation tag. This member recommendation tag includes a "click to play" tag, a first video viewing duration tag, and a second video viewing duration tag. The first video viewing duration tag indicates that the viewing duration of the first target video exceeds a preset percentage of the total video duration. The second video viewing duration tag indicates that the viewing duration of the second target video equals the total video duration. The first target video and the second target video can be any two different videos; or...

[0083] When the target identifier's identifier category is non-member, the non-member recommendation tag is used as the target recommendation tag. The non-member recommendation tag includes a click-to-play tag, a third video viewing duration tag, and an order page click tag. The third video viewing duration tag is used to indicate that the viewing duration of the third target video has reached a preset duration. The order page click tag is used to indicate that a non-member clicks on the member order page to contribute to the member recharge and convert to a member. The third target video and the first target video and the second target video can be any video that is the same or different.

[0084] In one embodiment, the information flow determination module 330 is specifically used for:

[0085] The recommendation value corresponding to the candidate information stream is obtained by weighted summing of the correlation between the candidate information stream and each of the behavior tags.

[0086] The candidate information streams whose recommendation values ​​meet the recommendation criteria are taken as the information streams to be recommended, wherein the recommendation criteria include a recommendation value greater than or equal to a preset value, and / or, the recommendation value is located before or after a preset rank in the recommendation value ranking.

[0087] In one embodiment, the information flow determination module 330 is specifically used for:

[0088] Based on the attribute description vector corresponding to the target attribute information, a corresponding feature weight vector is constructed, wherein the feature weight vector is used to indicate the preference of the target identifier for each attribute feature;

[0089] According to the target recommendation label, the embedding mapping results of the corresponding information description vectors and feature weight vectors of each candidate information stream are classified to obtain the correlation degree between each candidate information stream and the target recommendation label.

[0090] In one embodiment, the information flow determination module 330 is specifically used for:

[0091] The target attribute information is mapped and transformed to the corresponding attribute description vector, wherein the attribute description vector is composed of feature vectors corresponding to multiple attribute features;

[0092] Obtain preset key features, wherein the preset key features include at least one attribute feature among the plurality of attribute features;

[0093] A first weight vector is formed based on the feature vectors corresponding to the preset key features in the attribute description vector;

[0094] The feature weight vector is generated by combining the attribute description vector with the first weight vector.

[0095] In one embodiment, the information flow determination module 330 is specifically used for:

[0096] The first weight vector and the attribute description vector are subjected to basic operations to obtain the second weight vector, wherein the basic operations include multiplication, subtraction and addition;

[0097] The combined vector formed by concatenating the second weight vector and the attribute description vector is used as the feature weight vector.

[0098] In one embodiment, the information flow determination module 330 is specifically used for:

[0099] Based on the material description vector and / or identity description vector corresponding to the target attribute information, a corresponding feature weight vector is constructed, wherein the attribute description vector includes the material description vector and / or identity description vector.

[0100] Figure 4 An internal structural diagram of a computer device in one embodiment is shown. Specifically, this computer device may be... Figure 1 Server 120 in the middle. For example... Figure 4 As shown, the computer device includes a processor, memory, network interface, input device, and display screen connected via a system bus. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and may also store computer programs. When executed by the processor, these programs enable the processor to implement an information flow recommendation method. The internal memory may also store computer programs, which, when executed by the processor, enable the processor to implement the information flow recommendation method. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0101] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0102] In one embodiment, the information flow recommendation device provided in this application can be implemented as a computer program, and the computer program can be implemented as follows: Figure 4 It runs on the computer device shown. The computer device's memory can store the various program modules that make up the information flow recommendation device, for example, Figure 3 The diagram shows the acquisition module 310, target determination module 320, information flow determination module 330, and push module 340. The computer program comprised of these modules causes the processor to execute the steps in the information flow recommendation methods of the various embodiments of this application described in this specification.

[0103] Figure 4 The computer device shown can be used as follows Figure 3 The acquisition module 310 in the information flow recommendation device shown acquires target attribute information corresponding to the target identifier. The target attribute information includes the identifier category of the target identifier, which includes members and non-members. The computer device can use the target determination module 320 to determine corresponding target recommendation tags based on the identifier category of the target identifier. These target recommendation tags are either member recommendation tags or non-member recommendation tags. The member recommendation tags are used to guide members to contribute video viewing time, while the non-member recommendation tags are used to guide non-members to contribute to member subscriptions. The computer device can use the information flow determination module 330 to determine the information flow to be recommended based on the correlation between each candidate information flow and the target recommendation tag. The computer device can use the push module 340 to push the information flow to be recommended to the terminal corresponding to the target identifier.

[0104] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in any of the above embodiments.

[0105] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the above embodiments.

[0106] 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. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0107] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0108] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An information flow recommendation method, characterized in that, The method includes: Obtain target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members; The corresponding target recommendation tag is determined according to the identification category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute member recharge to convert to member status. Based on the attribute description vector corresponding to the target attribute information, a corresponding feature weight vector is constructed. The attribute description vector includes feature vectors corresponding to multiple attribute features, and the feature weight vector includes the target identifier's preference for each attribute feature. The attribute description vector is converted into a feature weight vector based on the Attention mechanism. According to the target recommendation label, the embedding mapping results of the corresponding information description vectors and the feature weight vectors of each candidate information stream are classified to obtain the correlation degree between each candidate information stream and the target recommendation label; The information stream to be recommended is determined based on the correlation between each candidate information stream and the target recommendation tag; The recommended information stream is pushed to the terminal corresponding to the target identifier.

2. The method according to claim 1, characterized in that, The step of determining the corresponding target recommendation tag based on the identifier category of the target identifier includes: When the target identifier's identifier category is "member," a "member recommendation" tag is used as the target recommendation tag. This member recommendation tag includes a "click to play" tag, a first video viewing duration tag, and a second video viewing duration tag. The first video viewing duration tag indicates that the viewing duration of the first target video exceeds a preset percentage of the total video duration. The second video viewing duration tag indicates that the viewing duration of the second target video equals the total video duration. The first target video and the second target video can be any two different videos; or... When the target identifier's identifier category is non-member, the non-member recommendation tag is used as the target recommendation tag. The non-member recommendation tag includes a click-to-play tag, a third video viewing duration tag, and an order page click tag. The third video viewing duration tag is used to indicate that the viewing duration of the third target video has reached a preset duration. The order page click tag is used to indicate that a non-member clicks on the member order page to contribute to the member recharge and convert to a member. The third target video and the first target video and the second target video can be any video that is the same or different.

3. The method according to claim 2, characterized in that, The target recommendation tag includes multiple behavioral tags, which are click-to-play tags, first video viewing duration tags, second video viewing duration tags, third video viewing duration tags, or order page click tags. The step of determining the information stream to be recommended based on the correlation between each candidate information stream and the target recommendation tag includes: The recommendation value corresponding to the candidate information stream is obtained by weighted summing of the correlation between the candidate information stream and each of the behavior tags. The candidate information streams whose recommendation values ​​meet the recommendation criteria are taken as the information streams to be recommended, wherein the recommendation criteria include a recommendation value greater than or equal to a preset value, and / or, the recommendation value is located before or after a preset rank in the recommendation value ranking.

4. The method according to claim 1, characterized in that, The construction of the corresponding feature weight vector based on the attribute description vector corresponding to the target attribute information includes: The target attribute information is mapped and transformed to the corresponding attribute description vector, wherein the attribute description vector is composed of feature vectors corresponding to multiple attribute features; Obtain preset key features, wherein the preset key features include at least one attribute feature among the plurality of attribute features; A first weight vector is formed based on the feature vectors corresponding to the preset key features in the attribute description vector; The feature weight vector is generated by combining the attribute description vector with the first weight vector.

5. The method according to claim 4, characterized in that, The step of generating the feature weight vector by combining the attribute description vector with the first weight vector includes: The first weight vector and the attribute description vector are subjected to basic operations to obtain the second weight vector, wherein the basic operations include multiplication, subtraction and addition; The combined vector formed by concatenating the second weight vector and the attribute description vector is used as the feature weight vector.

6. The method according to claim 1, characterized in that, The construction of the corresponding feature weight vector based on the attribute description vector corresponding to the target attribute information includes: Based on the material description vector and / or identity description vector corresponding to the target attribute information, a corresponding feature weight vector is constructed, wherein the attribute description vector includes the material description vector and / or identity description vector.

7. An information flow recommendation device, characterized in that, For implementing the information flow recommendation method as described in claim 1, the apparatus includes: The acquisition module is used to acquire target attribute information corresponding to the target identifier, wherein the target attribute information includes the identifier category of the target identifier, and the identifier category includes members and non-members; The target determination module is used to determine the corresponding target recommendation tag according to the identifier category of the target identifier, wherein the target recommendation tag is a member recommendation tag or a non-member recommendation tag. The member recommendation tag is used to guide members to contribute video viewing time, and the non-member recommendation tag is used to guide non-members to contribute to the conversion of members to members through recharge. The information flow determination module is used to construct corresponding feature weight vectors based on the attribute description vectors corresponding to the target attribute information. The attribute description vectors include feature vectors corresponding to multiple attribute features, and the feature weight vectors include the target identifier's preference for each attribute feature. The attribute description vectors are converted to feature weight vectors based on an attention mechanism. According to the target recommendation tag, the embedding mapping results of the information description vectors and feature weight vectors of each candidate information flow are classified to obtain the correlation between each candidate information flow and the target recommendation tag. Based on the correlation between each candidate information flow and the target recommendation tag, the information flow to be recommended is determined. The push module is used to push the information stream to be recommended to the terminal corresponding to the target identifier.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

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

Images