Information processing method and device, electronic equipment and storage medium

Abstract

This application discloses an information processing method, apparatus, electronic device, and storage medium. Embodiments of this invention can be applied to various scenarios such as cloud technology, artificial intelligence, smart mobility, and assisted driving. The method includes: obtaining rating thresholds for target information from multiple users; determining the rating of the information category corresponding to the target information from an information rating matrix; if a user's target rating is higher than their rating threshold, designating that user as the target user and sending the target information to that user. By integrating the information category of the newly added information into the existing rating matrix, the ratings in the obtained information rating matrix accurately reflect the probability of the newly added information being selected by the user, thereby increasing the accuracy of identifying the target user and reducing the occurrence of sending target information to users who are not interested in the target information. This achieves the effect of precise delivery of target information, improves the efficiency of information transmission, and enhances the user experience.

Description

Technical Field

[0001] This application relates to the field of Internet information processing technology, and more specifically, to an information processing method, apparatus, electronic device, and storage medium. Background Technology

[0002] Intelligent selection (recommendation) service refers to a service that uses big data and artificial intelligence technologies, combined with the accumulation of knowledge from multiple industry sectors, to select information that users may like. For example, it can select products that users may like based on their browsing and purchase history.

[0003] Currently, in the application scenarios of intelligent selection services, historical data of multiple users regarding existing information can be obtained. Then, based on this historical data, the selection probability of each user regarding existing information can be determined, and users with higher selection probabilities can be selected from multiple users as target users. The existing information is then sent to the target users.

[0004] However, new information lacks historical data such as users' browsing history, click history, purchase history, and collection history. Using the above methods, it is impossible to obtain accurate target users, which may result in new information being sent to users who do not pay attention to it, leading to low information delivery efficiency and poor user experience. Summary of the Invention

[0005] In view of this, embodiments of this application propose an information processing method, apparatus, electronic device, and storage medium.

[0006] In a first aspect, embodiments of this application provide an information processing method, the method comprising: obtaining rating thresholds for target information for each of multiple users, wherein the target information is at least one of newly added information, and the information category of the newly added information belongs to at least one of multiple preset information categories; determining the rating of the information category corresponding to the target information from an information rating matrix as the target rating, wherein the information rating matrix includes the ratings of multiple users for each preset information category, the target rating includes the target ratings of multiple users for the target information, the information rating matrix is ​​obtained based on the existing rating matrix of existing information and the new category rating matrix of newly added information, the existing rating matrix includes the probability that existing information under each preset information category is selected by each user, and the new category rating matrix includes the probability that the information category of each newly added information belongs to each preset information category; if a user's target rating is higher than the user's rating threshold, the user is designated as the target user, and the target information is sent to the target user.

[0007] Secondly, embodiments of this application provide an information processing apparatus, comprising: an acquisition module, configured to acquire rating thresholds for target information from multiple users, wherein the target information is at least one of newly added information, and the information category of the newly added information belongs to at least one of multiple preset information categories; a determination module, configured to determine the rating of the information category corresponding to the target information from an information rating matrix, wherein the information rating matrix includes the ratings of multiple users for each preset information category, the target rating includes the target ratings of multiple users for the target information, the information rating matrix is ​​obtained based on the existing rating matrix of existing information and the new category rating matrix of newly added information, the existing rating matrix includes the probability that existing information under each preset information category is selected by each user, and the new category rating matrix includes the probability that the information category of each newly added information belongs to each preset information category; and a sending module, configured to, if a user's target rating is higher than the user's rating threshold, designate the user as the target user and send the target information to the target user.

[0008] Optionally, the device further includes an information rating matrix acquisition module, which performs singular value decomposition on an existing rating matrix to obtain an information interest matrix, a user interest matrix, and a covariance matrix; obtains a new information interest matrix based on the information interest matrix and the newly added category rating matrix; and obtains an information rating matrix based on the covariance matrix, the user interest matrix, and the new information interest matrix.

[0009] Optionally, the information rating matrix acquisition module is also used to calculate the product of the information interest matrix and the inverse of the new category rating matrix as the conditional new category rating matrix; and to calculate the product of the conditional new category rating matrix and the new category rating matrix as the new information interest matrix.

[0010] Optionally, the information rating matrix acquisition module is also used to calculate the product of the covariance matrix, the transpose of the user interest matrix, and the new information interest matrix, as the information rating matrix.

[0011] Optionally, the acquisition module is further configured to determine the information category of the target information; determine the average score of the information category corresponding to the target information from the average score matrix as a score threshold. The average score matrix includes the average scores of multiple users for each preset information category. The average score in the i-th row and j-th column of the average score matrix represents the attention of the i-th user to the information in the j-th preset information category. i is any integer value between 1 and N, j is any integer value between 1 and M, N is the total number of multiple users, and M is the total number of multiple preset information categories.

[0012] Optionally, the device further includes an average rating matrix acquisition module, used to acquire target sample features for each of the multiple users regarding the target sample information, wherein the target sample information is determined from the sample information and does not include new information; process the target sample features through an information rating model to obtain the predicted ratings for each of the multiple users regarding each target sample information, wherein the predicted rating for each user regarding each target sample information represents the probability that the target sample information is selected by the user; and obtain the average rating matrix based on the predicted ratings for each of the multiple users regarding each target sample information.

[0013] Optionally, the average rating matrix acquisition module is further configured to: determine that the target sample information has been selected by the user if the user's predicted rating for the target sample information reaches the user's target rating threshold for the target sample information; calculate the ratio of the number of target sample information selected by the nth user under the mth preset information category to the total number of all target sample information under the mth preset information category, as the probability that the target sample information under the mth preset information category is selected by the nth user, where m takes integer values ​​from 1 to M, n takes integer values ​​from 1 to N, M is the total number of multiple preset information categories, and N is the total number of multiple users; and construct an average rating matrix based on the probability that the target sample information under each preset information category is selected by each user.

[0014] Optionally, the device further includes a model training module for acquiring sample category features for multiple preset information categories, wherein the sample category features do not include features of newly added information; dividing the sample category features into sparse features and dense features; processing the sparse features through a deep network layer in the initial model to obtain a first feature; processing the dense features through a linear network layer in the initial model to obtain a second feature; fusing the first feature and the second feature to obtain a fused feature; and training the initial model based on the fused feature to obtain an information scoring model.

[0015] Optionally, the device further includes a new category scoring matrix acquisition module, used to determine the proportion of new information in each preset information category based on the information attributes of each new information and the total number of information in each preset information category, and to use the proportion of new information in each preset information category as the probability that the information category of each new information belongs to each preset information category; and to construct a new category scoring matrix based on the probability that the information category of each new information belongs to each preset information category.

[0016] Optionally, the device further includes an existing rating matrix acquisition module, used to acquire the selection results of multiple users for existing information, whereby each user's selection result includes whether existing information was selected by the user or not. The module calculates the ratio of the number of existing information items selected by the q-th user under the p-th preset information category to the total number of all existing information items under the p-th preset information category, as the probability that existing information under the p-th preset information category is selected by the q-th user. p ranges from 1 to M, q ranges from 1 to N, M is the total number of multiple preset information categories, and N is the total number of multiple users. Based on the probability that existing information under each preset information category is selected by each user, an existing rating matrix is ​​constructed.

[0017] Thirdly, embodiments of this application provide an electronic device, including a processor and a memory; one or more programs are stored in the memory and configured to be executed by the processor to implement the above-described method.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium storing program code, wherein the above-described method is executed when the program code is run by a processor.

[0019] Fifthly, embodiments of this application provide a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the method described above.

[0020] This application provides an information processing method, apparatus, electronic device, and storage medium. Since the existing scoring matrix includes the probability of each user selecting existing information under each preset information category, and the new category scoring matrix includes the probability that the information category of each new piece of information belongs to each preset information category, an information scoring matrix is ​​obtained by combining the existing scoring matrix and the new category scoring matrix. This integrates the information categories of new information into the existing scoring matrix, ensuring that the scores in the obtained information scoring matrix accurately reflect the probability of users selecting new information. This results in a higher accuracy rate for the target scores of the identified target information, and consequently, a higher accuracy rate for identifying target users. This reduces the occurrence of sending target information to users who do not care about the target information, achieving the effect of accurate target information delivery, improving information transmission efficiency, and enhancing user experience. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application;

[0023] Figure 2 A flowchart of an information processing method according to an embodiment of this application is shown;

[0024] Figure 3 A flowchart of an information processing method according to yet another embodiment of this application is shown;

[0025] Figure 4 A flowchart of an information processing method according to another embodiment of this application is shown;

[0026] Figure 5 A flowchart of a training method for an information scoring model in an embodiment of this application is shown;

[0027] Figure 6 A flowchart illustrating another training method for the information scoring model in an embodiment of this application is shown;

[0028] Figure 7 A block diagram of an information processing apparatus according to one embodiment of this application is shown;

[0029] Figure 8 A structural block diagram of an electronic device for performing an information processing method according to an embodiment of this application is shown. Detailed Implementation

[0030] 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, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0031] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0033] It should be noted that "multiple" in this article refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0034] Big data refers to data sets that cannot be captured, managed, and processed within a certain timeframe using conventional software tools. It represents massive, rapidly growing, and diverse information assets that require new processing models to achieve stronger decision-making, insightful discovery, and process optimization capabilities. With the advent of the cloud era, big data has attracted increasing attention. Big data requires specialized technologies to effectively process large amounts of data within a tolerable timeframe. Technologies suitable for big data include massively parallel processing databases, data mining, distributed file systems, distributed databases, cloud computing platforms, the internet, and scalable storage systems.

[0035] It should be noted that, in this embodiment of the application, the acquisition of user data such as browsing history, click history, purchase history, and collection history requires the user's permission or consent, and the collection, use, processing, and storage of user data such as browsing history, click history, purchase history, and collection history must comply with the regulations of the region.

[0036] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application. For example... Figure 1 As shown, this application scenario includes terminal 101 and server 102, which are connected via wired or wireless networks. Terminal 101 can be a mobile phone, computer, smart voice interaction device, smart home appliance, vehicle terminal, aircraft, etc., but is not limited to these.

[0037] Server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Server 102 can be used to provide services for the applications running on terminal 101.

[0038] The embodiments of this invention can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart mobility, and assisted driving. For example, it can be used in scenarios involving the promotion of new mobility products.

[0039] The terminal 101 can acquire user data from the user side and send the user data from the user side to the server 102. The server 102 processes the acquired information data from the information side and user data from the user side (for example, using the information data from the information side and user data from the user side as input to the training samples to train an initial model and obtain an information selection model).

[0040] Information can refer to physical or virtual items used for browsing, purchasing, and use. For example, information can be physical goods, multimedia information, web links, and travel service products.

[0041] New information can refer to information that is not yet in use or is about to be put into use, such as newly launched products, songs, and new travel service products. Conversely, existing information can refer to information that is already in use, such as already launched products, songs, and travel service products.

[0042] For example, all songs released in the xx Music app before October 1, 2020 are considered existing information as of October 1, 2020, and new unreleased songs acquired by the xx Music app on October 1, 2020 can be considered new information as of October 1, 2020.

[0043] Server 102 can also obtain target information input by specific personnel (such as technicians responsible for publishing new information or technicians maintaining information), and execute the information processing method of this application to determine the target user among multiple users and determine the terminal 101 corresponding to the target user. Then, server 102 sends the target information to the terminal 101 corresponding to the target user.

[0044] Collaborative filtering (CF) is a technique used to analyze user interests. It involves finding users with similar interests within a user group, combining their evaluations of a particular piece of information, and then using this combined information to predict the user group's level of preference for that information.

[0045] The paired collaborative filtering (CF) selection algorithm first constructs a collaborative filtering algorithm based on existing information, and then, under the conditions of this existing collaborative filtering algorithm, constructs a selection algorithm model for new information. The paired CF selection algorithm achieves effective selection of new information to a certain extent.

[0046] The inventors discovered through research that existing information selection methods lack information such as user browsing history, click history, purchase history, and favorites history when adding new information. Therefore, new information cannot correspond to an actual existing information-side rating matrix. The proposed solution involves fitting the information-side rating matrix to all information, including new information, to obtain a fitted information-side rating matrix. Then, based on the user-side rating matrix and the fitted information-side rating matrix, a final information selection rating matrix is ​​obtained. This information selection rating matrix is ​​used to identify the target user among multiple users.

[0047] Although the paired CF selection algorithm achieves effective selection of new information to a certain extent, since each new piece of information does not have rating records, click records, purchase records, etc., when the method is used to fit the information-side rating matrix, the fitting value of the new information in the information-side rating matrix is ​​zero. This results in the new information having a zero rating in the information selection rating matrix determined based on the fitted information-side rating matrix, leading to no solution after the information selection rating matrix is ​​decomposed, and thus the accuracy of the information selection rating matrix remains low.

[0048] Based on this, the inventors have proposed the information processing method, apparatus, electronic device and storage medium provided in this application. Through the method of the embodiments of this application, the scores in the obtained information scoring matrix can accurately reflect the probability that the new information is selected by the user, thereby making the accuracy of the target score of the determined target information higher, and thus making the accuracy of the determined target user higher, reducing the occurrence of target information being sent to users who do not pay attention to the target information, achieving the effect of accurate delivery of target information and improving user experience.

[0049] Please see Figure 2 , Figure 2 This application illustrates a flowchart of an information processing method according to an embodiment of the present application. This method can be applied to an electronic device, which may be... Figure 1 Server 102 in the middle, the method includes:

[0050] S110. Obtain the rating thresholds for target information for each of multiple users, where the target information is at least one of the newly added information and the information category of the newly added information belongs to at least one of multiple preset information categories.

[0051] In this application, the information category can refer to the type, attribute, or category of information. For example, when the information is a song, the information category can be instrumental music, pop music, rock, blues, and classical music, etc. Similarly, when the information is a product, the information category can be clothing, electronic products, virtual products, books, and food, etc.

[0052] Preset information categories can refer to all categories of information, or they can be selected from all categories of information. For example, when the information is about goods, the information categories include clothing, electronic products, virtual products, books, food, bags, and jewelry. The preset information categories can be clothing, electronic products, books, food, bags, and jewelry.

[0053] The target information can include one or more. Multiple users may have different rating thresholds for the same target information, or they may have the same rating threshold. For example, if the target information is a top (A), and there are 10 users, each with a different rating threshold for top A. Conversely, if the target information is a handbag (B), and there are 5 users, each with a different rating threshold for handbag B.

[0054] The rating thresholds for the same user can be different or the same for different target information. For example, the rating threshold for target information a1 by the same user b1 is different from the rating threshold for target information a2, while the rating threshold for target information a3 by the same user b2 is the same as the rating threshold for target information a4.

[0055] As one implementation method, based on demand, multiple users can set their own rating thresholds for each preset information category, obtain the preset information category that is the same as the target information category as the target preset information category, and use the rating thresholds of multiple users under the target preset information category as the rating thresholds of multiple users for the target information.

[0056] For example, if the information is a product, and the preset information categories are clothing, electronics, books, food, bags, and jewelry, and there are 5 users, the set rating thresholds include 5 rating thresholds for clothing, 5 rating thresholds for electronics, 5 rating thresholds for books, 5 rating thresholds for food, 5 rating thresholds for bags, and 5 rating thresholds for jewelry for the 5 users; when the target information is top A, the information category of top A is determined to be clothing, the preset information category of clothing is used as the target preset information category, and the 5 rating thresholds under the clothing category for the 5 users are used as the rating thresholds for top A for each of the 5 users.

[0057] As another implementation method, an information selection model can be used to predict existing information. Based on the prediction results of the existing information, a rating matrix can be constructed. The rating matrix includes the ratings of multiple users for each preset information category. The information category corresponding to the target information is determined from the rating matrix as the target preset information category, and the rating corresponding to the target preset information category is used as the rating threshold.

[0058] For example, the information is about goods, with preset information categories including clothing, electronics, books, food, bags, and jewelry. There are six users, and a 6×6 rating matrix is ​​constructed. Each row of the matrix represents the rating of each user for each of the six preset information categories, and each column represents the ratings of each of the six users for each preset information category. When the target information is mobile phone C, the information category for mobile phone C is determined to be electronics. The preset information category "electronics" is then used as the target preset information category, and the six ratings from the six users for "electronics" are used as the rating thresholds for mobile phone C.

[0059] S120. Determine the score of the information category corresponding to the target information from the information scoring matrix as the target score. The information scoring matrix includes the scores of multiple users for each preset information category. The target score includes the target scores of multiple users for the target information. The information scoring matrix is ​​obtained based on the existing scoring matrix of existing information and the new category scoring matrix of new information. The existing scoring matrix includes the probability that existing information under each preset information category is selected by each user. The new category scoring matrix includes the probability that the information category of each new information belongs to each preset information category.

[0060] In this embodiment, an existing rating matrix can be constructed based on existing information, and a new category rating matrix can be constructed based on newly added information. Then, based on the existing rating matrix and the new category rating matrix, an information rating matrix is ​​obtained. The information rating matrix includes the ratings of multiple users for each preset information category. For example, if there are 7 preset information categories and 6 users, the constructed information rating matrix is ​​a 6×7 matrix. Each row of the information rating matrix is ​​the rating of each user for the 7 different preset information categories, and each column of the information rating matrix is ​​the rating of each of the 6 users for each preset information category.

[0061] The existing rating matrix includes the probability of each user selecting existing information under each preset information category. For example, if there are 7 preset information categories and multiple users, each user has 6 categories, the constructed existing rating matrix is ​​a 6×7 matrix. Each row of the existing rating matrix contains the probability of each user for each of the 7 preset information categories, and each column contains the probability of each user for each of the 6 preset information categories.

[0062] As one implementation method, the method for obtaining the existing rating matrix includes: obtaining the selection results of multiple users for existing information, where each user's selection result includes whether the existing information was selected by the user or not; calculating the ratio of the number of existing information selected by the qth user under the p-th preset information category to the total number of all existing information under the p-th preset information category, as the probability that the existing information under the p-th preset information category is selected by the qth user, where p takes values ​​from 1 to M, q takes values ​​from 1 to N, M is the total number of multiple preset information categories, and N is the total number of multiple users; and constructing the existing rating matrix based on the probability that the existing information under each preset information category is selected by each user.

[0063] As described above, the existing information is publicly available information. The existing information may be selected by the user or it may not be selected. Therefore, each user has a selection result for each piece of existing information.

[0064] For example, there are 4 preset information categories, and multiple users are categorized into 3 categories. Each user corresponds to 100 existing information items under each preset information category. For user 1, the selected existing information items under preset information category 1 are 20, under preset information category 2 are 30, under preset information category 3 are 25, and under preset information category 4 are 28. For user 2, the selected existing information items under preset information category 1 are 21, under preset information category 2 are 31, under preset information category 3 are 26, and under preset information category 4 are 29. For user 3, the selected existing information items under preset information category 1 are 22, under preset information category 2 are 32, under preset information category 3 are 27, and under preset information category 4 are 30. Based on the above data, the constructed existing rating matrix is ​​as follows:

[0065]

[0066] Each row, from left to right, represents the probabilities corresponding to preset information category 1, preset information category 2, preset information category 3, and preset information category 4. Each column, from top to bottom, represents the probabilities corresponding to user 1, user 2, and user 3.

[0067] The new category scoring matrix includes the probability that each new piece of information belongs to each preset information category. For example, if there are 8 new pieces of information and 10 preset information categories, the resulting new category scoring matrix is ​​an 8×10 matrix. Each row of the new category scoring matrix includes the probability corresponding to each of the 10 preset information categories, and each column of the new category scoring matrix includes the probability corresponding to each of the 8 new pieces of information.

[0068] As one implementation method, the method for obtaining the new category scoring matrix includes: determining the proportion of new information in each preset information category based on the information attributes of each new piece of information and the total number of information in each preset information category; using the proportion of new information in each preset information category as the probability that the information category of each new piece of information belongs to each preset information category; and constructing the new category scoring matrix based on the probability that the information category of each new piece of information belongs to each preset information category.

[0069] For example, there are 3 preset information categories: preset information category 1 has a total of 10 pieces of information, preset information category 2 has a total of 5 pieces of information, and preset information category 3 has a total of 20 pieces of information. There are 3 new information categories: new information 1 belongs to preset information category 1, new information 2 belongs to preset information category 2, and new information 3 belongs to preset information category 2. The constructed new category matrix is ​​as follows:

[0070]

[0071] Each row, from left to right, represents the probabilities of preset information category 1, preset information category 2, and preset information category 3, while each column, from top to bottom, represents the probabilities of newly added information 1, newly added information 2, and newly added information 3.

[0072] After obtaining the existing rating matrix and the new category matrix, the information rating matrix is ​​determined based on the existing rating matrix and the new category matrix. Then, the preset information category corresponding to the target information is determined from the information rating matrix as the selected preset information category. The ratings of multiple users corresponding to the selected preset information category are taken as the target ratings of each user.

[0073] For example, the information is about goods, with preset information categories including clothing, electronics, books, food, bags, and jewelry. There are six users, and the information rating matrix is ​​a 6x6 matrix. Each row of the matrix represents each user's rating for the six preset information categories, and each column represents the ratings of each of the six users for each preset information category. When the target information is the book "Living," the information category for "Living" is determined to be "books." The preset information category "books" is then selected, and the six ratings from the six users for "books" are taken as the target ratings for "Living."

[0074] S130. If a user's target rating is higher than the user's rating threshold, the user is designated as the target user, and the target information is sent to the target user.

[0075] For each user, determine whether the user's target score is higher than the score threshold. If so, the user is considered a target user. If not, the user is not considered a target user. Iterate through all users to identify all target users and send the target information to all target users.

[0076] As one implementation method, after identifying the target user, the user terminal corresponding to the target user is obtained, and then the target information is sent to the user terminal corresponding to the target user, thereby realizing the sending of the target information to the target user.

[0077] In this embodiment, since the existing rating matrix includes the probability of each user selecting existing information under each preset information category, and the new category rating matrix includes the probability that the information category of each new information belongs to each preset information category, an information rating matrix is ​​obtained by combining the existing rating matrix and the new category rating matrix. This allows the information category of the new information to be integrated into the existing rating matrix, so that the rating in the obtained information rating matrix can accurately reflect the probability of the new information being selected by the user. This results in a higher accuracy rate for the target rating of the identified target information, which in turn leads to a higher accuracy rate for identifying the target users. This reduces the occurrence of sending target information to users who do not care about the target information, achieving the effect of accurate delivery of target information, improving the efficiency of information transmission, and enhancing the user experience.

[0078] Please see Figure 3 , Figure 3 This application illustrates a flowchart of an information processing method according to another embodiment of the present application. This method can be applied to an electronic device, which may be... Figure 1 Server 102 in the middle, the method includes:

[0079] S210. Perform singular value decomposition on the existing rating matrix to obtain the information interest matrix, user interest matrix, and covariance matrix.

[0080] After obtaining the existing rating matrix, singular value decomposition is performed on it to obtain the information interest matrix, user interest matrix, and covariance matrix. The singular value decomposition process can be referred to Formula 1, which is as follows:

[0081] R y =U T ZI

[0082] Among them, R yGiven an existing rating matrix, U is the user interest matrix, I is the information interest matrix, and Z is the covariance matrix, where R... y Let U be an N×M matrix, Z be an L×L matrix, I be an L×M matrix, N be the total number of users, M be the total number of preset information categories, and L be the total number of newly added information.

[0083] S220. Based on the information interest matrix and the newly added category rating matrix, a new information interest matrix is ​​obtained.

[0084] After obtaining the information interest matrix through singular value decomposition based on the existing rating matrix, a new information interest matrix is ​​determined according to the information interest matrix and the newly added category rating matrix. The process of determining the new information interest matrix may include: calculating the product of the information interest matrix and the inverse of the newly added category rating matrix as the conditional newly added category rating matrix; and calculating the product of the conditional newly added category rating matrix and the newly added category rating matrix as the new information interest matrix.

[0085] As mentioned above, the process of determining the new information interest matrix can be expressed as Formula 2, which is as follows:

[0086]

[0087] Where C is the newly added category rating matrix, I′ is the new information interest matrix, P is the newly added category rating matrix under the condition, P means the probability that the information category of each newly added information belongs to each of the preset information categories under each preset information category condition, and C is an L×M matrix.

[0088] S230. Obtain the information rating matrix based on the covariance matrix, the user interest matrix, and the new information interest matrix.

[0089] After obtaining the covariance matrix and user interest matrix based on the existing rating matrix, a new information interest matrix determined by S220 is obtained. Based on the covariance matrix, user interest matrix, and new information interest matrix, the information rating matrix is ​​obtained.

[0090] As one implementation method, the information rating matrix acquisition method includes: calculating the product of the covariance matrix, the transpose of the user interest matrix, and the new information interest matrix, as the information rating matrix.

[0091] As mentioned above, the process of determining the information scoring matrix can be expressed as Formula 3, which is as follows:

[0092] R = U T Zi′

[0093] Where R is the information rating matrix, and R is an N×M matrix.

[0094] S240. Obtain the rating thresholds for target information for each of multiple users, where the target information is at least one of the newly added information and the information category of the newly added information belongs to at least one of multiple preset information categories.

[0095] S250. Determine the score of the information category corresponding to the target information from the information scoring matrix as the target score. The information scoring matrix includes the scores of multiple users for each preset information category. The target score includes the target scores of multiple users for the target information. The information scoring matrix is ​​obtained based on the existing scoring matrix of existing information and the new category scoring matrix of new information. The existing scoring matrix includes the probability that existing information under each preset information category is selected by each user. The new category scoring matrix includes the probability that the information category of each new information belongs to each preset information category.

[0096] S260. If a user's target rating is higher than the user's rating threshold, the user is designated as the target user, and the target information is sent to the target user.

[0097] The descriptions of S240-S260 are the same as those of S110-S130 above, and will not be repeated here.

[0098] In this embodiment, singular value decomposition is performed on the existing scoring matrix, and the information scoring matrix is ​​obtained based on the decomposition results. The values ​​in the information scoring matrix can accurately reflect the scoring of the newly added information, thereby achieving accurate scoring of the newly added information and improving the accuracy of the target scoring.

[0099] The novel paired CF selection algorithm constructed using the method described in this application, and the information scoring matrix obtained through this algorithm, improves the accuracy of obtaining the information scoring matrix. The novel paired CF selection algorithm refers to the algorithm obtained by combining Formula 1, Formula 2, and Formula 3. The novel paired CF selection algorithm is as follows:

[0100]

[0101] Please see Figure 4 , Figure 4 This invention illustrates a flowchart of an information processing method according to another embodiment of the present application. This method can be applied to an electronic device, which may be... Figure 1 Server 102 in the middle, the method includes:

[0102] S310. Determine the information category of the target information; determine the average score of the information category corresponding to the target information from the average score matrix, as the score threshold. The average score matrix includes the average score of multiple users for each preset information category. The average score in the i-th row and j-th column of the average score matrix represents the attention of the i-th user to the information in the j-th preset information category. i is any integer value between 1 and N, j is any integer value between 1 and M, N is the total number of multiple users, and M is the total number of multiple preset information categories.

[0103] The preset information category corresponding to the target information is determined from the average rating matrix and used as the new target preset information category. The average rating of each user corresponding to the new target preset information category is used as the rating threshold for each user.

[0104] For example, the information is about goods, with preset information categories including clothing, electronics, books, food, and jewelry. There are five users, and the average rating matrix is ​​a 5x5 matrix. Each row of the matrix represents the average rating for each user across the five preset information categories, and each column represents the average rating for each of the five users within each preset information category. When the target information is a watch, the watch information category is determined as jewelry. This preset information category of jewelry is then defined as the new target preset information category. The average ratings for the five users under this new target information category of jewelry are used as the rating threshold for the watch.

[0105] As one implementation method, the method for obtaining the average rating matrix includes: obtaining target sample features for each of multiple users regarding target sample information, wherein the target sample information is determined from sample information and does not include new information; processing the target sample features through an information rating model to obtain the predicted ratings for each of the multiple users for each target sample information, wherein the predicted rating for each user for each target sample information represents the probability that the target sample information is selected by the user; and obtaining the average rating matrix based on the predicted ratings for each of the multiple users for each target sample information.

[0106] Sample features can refer to the characteristics of sample information. For each user, the sample features of each sample information can refer to the sum of the information data of that sample information and the user data of that user. The user data can refer to the user's clicks, favorites, comments, purchases, add-to-cart, payments, cancellations, and order cancellations related to that sample information. The information data of that sample information can refer to the click-through rate, comment rate, add-to-cart rate, and order cancellation rate of that sample information.

[0107] Target sample information can refer to a portion of the sample information identified in the sample information, or it can refer to all the sample information. Target sample features are the sample features of the identified target sample information.

[0108] It should be noted that newly added information does not include data such as clicks, favorites, comments, purchases, add-to-cart, payments, cancellations, or order refunds. Newly added information also does not include data such as click-through rate, comment rate, add-to-cart rate, or order refund rate. Therefore, the sample information does not include newly added information, and the determined target sample information also does not include newly added information.

[0109] Sample features can include the characteristics of sample information over different periods, such as daily, weekly, or monthly. For example, if the period is daily and the sample features are the characteristics of sample information from the most recent 20 days, then the target sample information can refer to the sample information from the most recent day.

[0110] Information scoring models can be used to determine the predicted score of target sample information. The information scoring model can be obtained by training a neural network model with training samples. The training samples used to train the information scoring model can include the sample features of the training sample information and the selection results of the training sample information. The neural network model processes the sample features of the training sample information to obtain the prediction result. Based on the prediction result and the selection results of the training sample information, the loss value is determined. Then, the neural network model is adjusted based on the loss value to obtain the information scoring model.

[0111] Input the target sample features of the x-th user for the y-th target sample information into the information scoring model to obtain the predicted score of the x-th user for the y-th target sample information output by the information scoring model. Here, the x-th user can be any one of multiple users, and the y-th target sample information can be any one of all target sample information. Iterate through multiple users and all target sample information to obtain the predicted scores of each user for each target sample information.

[0112] After obtaining the predicted scores from multiple users for each target sample, an average score matrix can be determined based on the obtained predicted scores.

[0113] As one implementation method, the method for obtaining the average rating matrix may include: if a user's predicted rating for target sample information reaches the user's target rating threshold for that target sample information, determining that the target sample information has been selected by the user; calculating the ratio of the number of target sample information selected by the nth user under the mth preset information category to the total number of all target sample information under the mth preset information category, as the probability that the target sample information under the mth preset information category is selected by the nth user, where m is an integer value from 1 to M, n is an integer value from 1 to N, M is the total number of multiple preset information categories, and N is the total number of multiple users; and constructing an average rating matrix based on the probability that the target sample information under each preset information category is selected by each user.

[0114] In this embodiment, the target scoring threshold for each user for each target sample information can be different. For example, in a scenario with 10 users and 10 target sample information, each user corresponds to one target scoring threshold for each target sample information, resulting in a total of 100 target scoring thresholds. Alternatively, the target scoring threshold for each user for the same target sample information can also be the same. For example, in a scenario with 10 users and 10 target sample information, multiple users correspond to one target scoring threshold for the same target sample information, resulting in a total of 10 target scoring thresholds for 10 target sample information. Furthermore, a single target scoring threshold can be set, regardless of the specific target sample information or user, and remains a fixed value.

[0115] By using the target scoring threshold and the predicted score, it is determined whether each target sample information is selected by each user. Then, the probability is calculated as the ratio of the number of target sample information selected by the nth user under the m-th preset information category to the total number of all target sample information under the m-th preset information category.

[0116] For example, if the number of target sample information selected by the nth user under the mth preset information category is 50, and the total number of all target sample information under the mth preset information category is 100, then the probability that the target sample information under the mth preset information category is selected by the nth user is 50 / 100 = 0.5.

[0117] Iterate through all users and multiple preset information categories to obtain the probability that each user selects the target sample information under each preset information category. Then, construct an average rating matrix based on the probability that each user selects the target sample information under each preset information category.

[0118] For example, there are 4 preset information categories and 3 users. Each user corresponds to 100 target sample information items under each preset information category. User 1 selects 32, 33, 34, and 35 target sample information items for preset information categories 1, 2, 3, and 4, respectively. User 2 selects 42, 43, 44, and 45 target sample information items for preset information categories 1, 2, 3, and 4, respectively. User 3 selects 52, 53, 54, and 55 target sample information items for preset information categories 1, 2, 3, and 4, respectively. In this case, the determined average rating matrix is ​​as follows:

[0119]

[0120] Each row, from left to right, represents the probabilities corresponding to preset information category 1, preset information category 2, preset information category 3, and preset information category 4. Each column, from top to bottom, represents the probabilities corresponding to user 1, user 2, and user 3.

[0121] S320. Determine the score of the information category corresponding to the target information from the information scoring matrix as the target score. The information scoring matrix includes the scores of multiple users for each preset information category. The target score includes the target scores of multiple users for the target information. The information scoring matrix is ​​obtained based on the existing scoring matrix of existing information and the new category scoring matrix of new information. The existing scoring matrix includes the probability that existing information under each preset information category is selected by each user. The new category scoring matrix includes the probability that the information category of each new information belongs to each preset information category.

[0122] S330. If a user's target rating is higher than the user's rating threshold, the user is designated as the target user, and the target information is sent to the target user.

[0123] The descriptions of S320 and S330 are the same as those of S120 and S130 above, and will not be repeated here.

[0124] In this embodiment, an average rating matrix is ​​determined based on the target sample information, and a rating threshold for the target information is obtained from the average rating matrix. The average rating matrix accurately reflects the user's attention to information under different preset information categories, thereby making the rating threshold more accurately reflect the user's attention to newly added information under the preset information categories. The rating threshold is more accurate, which makes the identified target users more accurate and improves the user experience.

[0125] Please see Figure 5 , Figure 5The flowchart illustrates a training method for an information scoring model in an embodiment of this application. This method can be applied to electronic devices, which may be... Figure 1 Server 102 in the middle, the method includes:

[0126] S410. Obtain sample category features for multiple preset information categories, where the sample category features do not include features of newly added information.

[0127] Category features can refer to the features of a preset information category. The category features of a preset information category can include information data and user data from multiple users for that preset information category. User data can include data from multiple users on clicks, favorites, comments, purchases, add-to-cart, payments, cancellations, and refunds for existing information under that preset information category. Information data can refer to data such as the average click-through rate, average comment rate, average add-to-cart rate, and average refund rate for existing information under that preset information category.

[0128] It should be noted that newly added information does not contain user data such as clicks, favorites, comments, purchases, add-to-cart, payments, cancellations, or refunds. Furthermore, it does not contain data such as average click-through rate, comment rate, add-to-cart rate, or refund rate. Therefore, the sample category features do not include features specific to newly added information.

[0129] Sample category features can refer to the category features used as training samples. For example, the category features can be divided into sample category features and test category features (the ratio of sample category features to test category features can be 8:2, that is, 80% of the data in the category features are used as sample category features). Sample category features are used to train the initial model, and test category features are used to test the performance of the initial model after training.

[0130] Categorical features can include categorical features from different periods, such as daily, weekly, or monthly. For example, with a daily period, the categorical features could be those from the last 20 days.

[0131] In one implementation, the sample category features within each period can be used as a batch of samples to train the initial model. The sample category features within different periods can be divided into different training batches. For example, if the period is daily and the sample category features are the sample category features of the most recent 20 days, the sample category features of the most recent 20 days can be divided into 20 batches of sample category features according to the days.

[0132] S420. Divide the sample category features into sparse features and dense features.

[0133] S430. The sparse features are processed through the deep network layers in the initial model to obtain the first feature.

[0134] S440. The dense features are processed through the linear network layer in the initial model to obtain the second feature.

[0135] The initial model may include deep network layers (DeepNet, which may include 5 deep network layers) for processing sparse features and linear network layers (Wide layers) for processing dense features.

[0136] During training, for each batch of sample category features, the sample category features are divided into sparse features and dense features. The sparse features are processed by the deep network layer in the initial model to obtain the first feature, where the first feature is the Embadding feature (this step realizes the transformation of discrete sparse features into continuous first features). The dense features are processed by the linear network layer in the initial model to obtain the second feature.

[0137] Iterate through the sample category features of all batches to obtain the first feature and second feature corresponding to the sample category features of all batches.

[0138] S450. The first feature and the second feature are fused to obtain the fused feature.

[0139] S460. Based on the fusion features, the initial model is trained to obtain the information scoring model.

[0140] For each batch of sample category features, after obtaining the first feature corresponding to the sparse feature and the second feature corresponding to the dense feature, the first feature and the second feature of the batch are fused to obtain the corresponding fused feature. The initial model is trained using the fused feature of the batch. By traversing the fused features corresponding to the sample category features of all batches, the initial model is trained based on all sample category features to obtain the information scoring model.

[0141] As one implementation, S460 may include: obtaining sample category labels corresponding to multiple preset information categories, wherein the sample category label of each preset information category indicates whether the preset information category is selected; processing the fusion features through an initial model to obtain the sample prediction result for each preset information category; determining the loss value based on the sample prediction result and sample category label corresponding to each preset information category; and training the initial model based on the loss value to obtain an information scoring model.

[0142] The sample category label for each preset information category indicates whether the preset information category is selected. When there is already selected information under a preset information category, it is determined that the preset information category is selected, and the sample category label "selected" is added to the preset information category (the sample category label can also be in numerical form, for example, 1 indicates selection); when there is no already selected information under a preset information category, it is determined that the preset information category is not selected, and the sample category label "not selected" is added to the preset information category (the sample category label can also be in numerical form, for example, 0 indicates not selected).

[0143] When the information is a product, selecting information can refer to purchasing information; when the information is other non-product information, selecting information can refer to clicking, browsing, or using information.

[0144] For each batch of sample category features, after obtaining the first feature corresponding to the sparse feature and the second feature corresponding to the dense feature, the first feature and the second feature of the batch are fused to obtain the fused feature. Then, the fused feature corresponding to the sample category features of the batch is processed by the initial model to obtain the sample prediction result for each preset information category. Based on the sample prediction result and sample category label corresponding to each preset information category, the loss value is determined. The sample category features of all batches are traversed to obtain all the loss values. The initial model is trained with all the loss values ​​to obtain the information scoring model.

[0145] In this embodiment, the sparse features in the sample category features of multiple preset information categories are processed by the deep network layer of the initial model, and the dense features in the sample category features of multiple preset information categories are processed by the linear network layer of the initial model. Different processing methods are implemented for features of different categories, so that the initial model can extract more accurate information from the sample category features of multiple preset information categories, thereby improving the training accuracy of the initial model and thus improving the effect of the information scoring model obtained by training.

[0146] Please see Figure 6 , Figure 6 This document illustrates a flowchart of another training method for the information scoring model in an embodiment of this application. This method can be applied to electronic devices, which may be... Figure 1 Server 102 in the middle, the method includes:

[0147] S510. Obtain sample category features for multiple preset information categories, where the sample category features do not include features of newly added information; divide the sample category features into sparse features and dense features; process the sparse features through the deep network layer in the initial model to obtain the first feature; process the dense features through the linear network layer in the initial model to obtain the second feature; fuse the first feature and the second feature to obtain the fused feature.

[0148] S520. Obtain sample category labels corresponding to multiple preset information categories. The sample category label of each preset information category indicates whether the preset information category is selected. Process the fusion features through the initial model to obtain the sample prediction result of each preset information category. Determine the loss value based on the sample prediction result and sample category label corresponding to each preset information category.

[0149] The descriptions of S510-S520 are the same as those of S410-S460 above, and will not be repeated here.

[0150] S530. Train the initial model based on the loss value to obtain the trained initial model.

[0151] The sample category features corresponding to each batch of preset information categories are used to obtain the sample prediction results corresponding to each preset information category. Based on the sample prediction results and sample category labels corresponding to each preset information category, the loss value is determined. The sample category features of all batches are traversed to obtain all the loss values. The initial model is trained using all the loss values ​​to obtain the trained initial model.

[0152] S540. Obtain test category features and test category labels for multiple preset information categories. The test category features do not include features of newly added information. The test category label for each preset information category indicates whether the preset information category is selected.

[0153] Test category features can refer to the category features used for testing. Test category features and test category labels are similar to those described above for sample category features and sample category labels, and will not be repeated here.

[0154] S550. The test category features are processed by the trained initial model to obtain the test prediction results for each preset information category.

[0155] The sparse features in the test category features can be processed by the deep network layer in the trained initial model to obtain the third feature, and the dense features in the test category features can be processed by the linear network layer in the trained initial model to obtain the fourth feature. Then, the third feature and the fourth feature are fused to obtain the fused test feature. The fused test feature is processed by the trained initial model to obtain the prediction result for each preset information category, which is used as the test prediction result for each preset information category.

[0156] S560. Determine the evaluation indicators based on the test prediction results and test category labels for each preset information category.

[0157] Evaluation metrics may include recall, precision, and AUC, where AUC stands for Area Under Curve, which means the area under the ROC curve and the coordinate axis.

[0158] Based on the test prediction results and test category labels for each preset information category, the recall, precision, and AUC values ​​of the initial model after training can be calculated and used as the query recall, precision, and AUC values ​​to be evaluated.

[0159] S570. When the evaluation indicators meet the preset conditions, the trained initial model will be used as the information scoring model.

[0160] The evaluation metric meeting the preset conditions can mean that the recall rate reaches the recall threshold, the precision rate reaches the precision threshold, and the AUC value reaches the AUC threshold. If any one of these occurs, the evaluation metric is determined to not meet the preset conditions. The recall threshold, precision threshold, and AUC threshold can be values ​​set based on requirements, and this application does not impose any restrictions.

[0161] The evaluation indicators meet the preset conditions, indicating that the initial model after training performs well. At this point, it is determined to be the information scoring model, which also performs well and has a high prediction accuracy.

[0162] If the evaluation indicators do not meet the preset conditions, new sample category features are obtained, and the initial model is trained again until the determined evaluation indicators meet the preset conditions, thus obtaining the information scoring model.

[0163] In this embodiment, the evaluation index of the initial model after training is determined, and the initial model after training is determined as the information scoring model only when the evaluation index meets the preset conditions, so that the obtained information scoring model has a better effect, thereby improving the prediction accuracy of the information scoring model.

[0164] To better understand the solution of this application, the information processing method of this application will be explained below with reference to an exemplary scenario.

[0165] In this exemplary scenario, there are 3 users, 4 preset information categories, and the target information is the information under preset information category 2. The sample category features include features from the last 10 days, and each sample category feature has a corresponding sample category label. The test category features include features from the last 10 days, and each test category feature has a corresponding test category label. The ratio of test category features to sample category features is 2:8. The target sample information is the sample information from the last day. The total number of target sample information under each preset information category is 100. There are 3 new pieces of information. The existing information under each preset information category includes 1000 pieces of information (the existing information includes the target sample information).

[0166] 1. Training process of information scoring model

[0167] The sample category features of the most recent 10 days are divided into 10 batches of sample category features, which are represented as X(t-9), X(t-8), X(t-7)...X(t-1) and X(t).

[0168] For each batch of sample category features, they are divided into sparse features Xu and dense features Xc. The sparse features Xu are processed through the deep network layer of the initial model to obtain the first feature, and the dense features Xc are processed through the linear network layer of the initial model to obtain the second feature. The first and second features are then fused to obtain the corresponding fused feature. The fused feature is processed by the initial model to obtain the corresponding sample prediction score. Based on the sample prediction score and the sample category label, the loss value of the sample category feature for that batch is determined, and the initial model is trained using this loss value. This process is repeated for all batches of sample category features to obtain all loss values, and the initial model is trained using all loss values ​​to obtain the trained initial model.

[0169] For each batch of test category features, they are divided into sparse features Xcu and dense features Xcc. The sparse features Xcu are processed by the deep network layer of the trained initial model to obtain the third feature, and the dense features Xcc are processed by the linear network layer of the trained initial model to obtain the fourth feature. The third and fourth features are then fused to obtain the corresponding fused test feature. This fused test feature is then processed by the trained initial model to obtain the corresponding test prediction score. All batches of test category features are iterated to obtain all test prediction scores. Based on all test prediction scores and test category labels, evaluation metrics are determined, which may include recall, precision, and AUC.

[0170] In this scenario, if the evaluation metrics meet the preset conditions, the trained initial model will be used as the information scoring model.

[0171] 2. Obtaining the existing rating matrix

[0172] For User 1, the selected existing information under preset information category 1, preset information category 2, preset information category 3, and preset information category 4 are 560, 600, 650, and 590 respectively. For User 2, the selected existing information under preset information category 1, preset information category 2, preset information category 3, and preset information category 4 are 550, 590, 640, and 580 respectively. For User 3, the selected existing information under preset information category 1, preset information category 2, preset information category 3, and preset information category 4 are 570, 610, 660, and 600 respectively. The determined existing rating matrix is ​​as follows:

[0173]

[0174] Each row, from left to right, represents the probabilities corresponding to preset information category 1, preset information category 2, preset information category 3, and preset information category 4. Each column, from top to bottom, represents the probabilities corresponding to user 1, user 2, and user 3.

[0175] 3. Added the ability to obtain the category rating matrix

[0176] The number of new information entries in preset information category 1 is 1, the number of new information entries in preset information category 2 is 1, and the number of new information entries in preset information category 3 is also 1. At this time, the total number of entries in the three preset information categories is 1001. The constructed new category matrix is ​​as follows:

[0177]

[0178] Each row, from left to right, represents the percentage of preset information category 1, preset information category 2, preset information category 3, and preset information category 4. Each column, from top to bottom, represents the percentage of newly added information 1, newly added information 2, and newly added information 3.

[0179] 4. Obtaining the average rating matrix

[0180] For each preset information category 1 corresponding to User 1, 100 target sample information are input into the initial model to obtain the preset evaluations for each of the 100 target sample information under that preset information category for that user. Among them, 50 reach the corresponding target score threshold, and the probability that the target sample information under that preset information category for that user is selected by the user is determined to be 0.5. Similarly, by iterating through all users and preset information categories, the probabilities of User 1 for preset information categories 1, 2, 3, and 4 are determined to be 0.5, 0.52, 0.49, and 0.55, respectively; the probabilities of User 2 for preset information categories 1, 2, 3, and 4 are determined to be 0.51, 0.53, 0.5, and 0.56, respectively; and the probabilities of User 3 for preset information categories 1, 2, 3, and 4 are determined to be 0.52, 0.54, 0.51, and 0.57, respectively. At this point, the determined average score matrix is:

[0181]

[0182] Each row, from left to right, represents the probabilities of preset information category 1, preset information category 2, preset information category 3, and preset information category 4. Each column, from top to bottom, represents the probabilities of newly added information 1, newly added information 2, and newly added information 3.

[0183] 5. Obtaining the Information Scoring Matrix

[0184] Perform singular value decomposition on the existing rating matrix to obtain the information interest matrix I, the user interest matrix U, and the covariance matrix Z; calculate the product of the information interest matrix I and the inverse of the new category rating matrix C as the conditional new category rating matrix P; calculate the product of the conditional new category rating matrix P and the new category rating matrix C as the new information interest matrix I′; calculate the product of the covariance matrix Z, the transpose of the user interest matrix U, and the new information interest matrix I′ as the information rating matrix R, where the information rating matrix is ​​also a 3×4 matrix.

[0185] In this scenario, the information scoring matrix is ​​as follows:

[0186]

[0187] 6. Identification of target users

[0188] The rating thresholds for users 1, 2, and 3 are determined to be 0.52, 0.53, and 0.54 respectively from the average rating matrix. The target ratings for users 1, 2, and 3 are determined to be 0.53, 0.52, and 0.55 respectively from the information rating matrix. Since the target ratings of users 1 and 3 are higher than the rating thresholds, users 1 and 3 are identified as target users, and target information is sent to the target users.

[0189] Please see Figure 7 , Figure 7 A block diagram of an information processing apparatus according to an embodiment of this application is shown. The apparatus 1100 includes:

[0190] The acquisition module 1110 is used to acquire the rating thresholds of multiple users for target information, where the target information is at least one of the newly added information and the information category of the newly added information belongs to at least one of multiple preset information categories.

[0191] The determining module 1120 is used to determine the score of the information category of the corresponding target information from the information scoring matrix. As the target score, the information scoring matrix includes the scores of multiple users for each preset information category. The target score includes the target scores of multiple users for the target information. The information scoring matrix is ​​obtained based on the existing scoring matrix of existing information and the new category scoring matrix of new information. The existing scoring matrix includes the probability that existing information under each preset information category is selected by each user. The new category scoring matrix includes the probability that the information category of each new information belongs to each preset information category.

[0192] The sending module 1130 is used to designate a user as a target user and send target information to the target user if the user's target score is higher than the user's score threshold.

[0193] Optionally, the device further includes an information rating matrix acquisition module, which performs singular value decomposition on an existing rating matrix to obtain an information interest matrix, a user interest matrix, and a covariance matrix; obtains a new information interest matrix based on the information interest matrix and the newly added category rating matrix; and obtains an information rating matrix based on the covariance matrix, the user interest matrix, and the new information interest matrix.

[0194] Optionally, the information rating matrix acquisition module is also used to calculate the product of the information interest matrix and the inverse of the new category rating matrix as the conditional new category rating matrix; and to calculate the product of the conditional new category rating matrix and the new category rating matrix as the new information interest matrix.

[0195] Optionally, the information rating matrix acquisition module is also used to calculate the product of the covariance matrix, the transpose of the user interest matrix, and the new information interest matrix, as the information rating matrix.

[0196] Optionally, the acquisition module 1110 is further configured to determine the information category of the target information; determine the average score of the information category corresponding to the target information from the average score matrix as a score threshold. The average score matrix includes the average score of multiple users for each preset information category. The average score in the i-th row and j-th column of the average score matrix represents the attention of the i-th user to the information in the j-th preset information category. i is any integer value between 1 and N, j is any integer value between 1 and M, N is the total number of multiple users, and M is the total number of multiple preset information categories.

[0197] Optionally, the device further includes an average rating matrix acquisition module, used to acquire target sample features for each of the multiple users regarding the target sample information, wherein the target sample information is determined from the sample information and does not include new information; process the target sample features through an information rating model to obtain the predicted ratings for each of the multiple users regarding each target sample information, wherein the predicted rating for each user regarding each target sample information represents the probability that the target sample information is selected by the user; and obtain the average rating matrix based on the predicted ratings for each of the multiple users regarding each target sample information.

[0198] Optionally, the average rating matrix acquisition module is further configured to: determine that the target sample information has been selected by the user if the user's predicted rating for the target sample information reaches the user's target rating threshold for the target sample information; calculate the ratio of the number of target sample information selected by the nth user under the mth preset information category to the total number of all target sample information under the mth preset information category, as the probability that the target sample information under the mth preset information category is selected by the nth user, where m takes integer values ​​from 1 to M, n takes integer values ​​from 1 to N, M is the total number of multiple preset information categories, and N is the total number of multiple users; and construct an average rating matrix based on the probability that the target sample information under each preset information category is selected by each user.

[0199] Optionally, the device further includes a model training module for acquiring sample category features for multiple preset information categories, wherein the sample category features do not include features of newly added information; dividing the sample category features into sparse features and dense features; processing the sparse features through a deep network layer in the initial model to obtain a first feature; processing the dense features through a linear network layer in the initial model to obtain a second feature; fusing the first feature and the second feature to obtain a fused feature; and training the initial model based on the fused feature to obtain an information scoring model.

[0200] Optionally, the device further includes a new category scoring matrix acquisition module, used to determine the proportion of new information in each preset information category based on the information attributes of each new information and the total number of information in each preset information category, and to use the proportion of new information in each preset information category as the probability that the information category of each new information belongs to each preset information category; and to construct a new category scoring matrix based on the probability that the information category of each new information belongs to each preset information category.

[0201] Optionally, the device further includes an existing rating matrix acquisition module, used to acquire the selection results of multiple users for existing information, whereby each user's selection result includes whether existing information was selected by the user or not. The module calculates the ratio of the number of existing information items selected by the q-th user under the p-th preset information category to the total number of all existing information items under the p-th preset information category, as the probability that existing information under the p-th preset information category is selected by the q-th user. p ranges from 1 to M, q ranges from 1 to N, M is the total number of multiple preset information categories, and N is the total number of multiple users. Based on the probability that existing information under each preset information category is selected by each user, an existing rating matrix is ​​constructed.

[0202] It should be noted that the device embodiments in this application correspond to the aforementioned method embodiments. The specific principles in the device embodiments can be found in the content of the aforementioned method embodiments, and will not be repeated here.

[0203] Figure 8 A structural block diagram of an electronic device for performing an information processing method according to an embodiment of this application is shown. The electronic device may be... Figure 1 Server 102, etc., should be noted that Figure 8 The computer system 1200 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0204] like Figure 8As shown, the computer system 1200 includes a Central Processing Unit (CPU) 1201, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on programs stored in Read-Only Memory (ROM) 1202 or programs loaded from storage portion 1208 into Random Access Memory (RAM) 1203. The RAM 1203 also stores various programs and data required for system operation. The CPU 1201, ROM 1202, and RAM 1203 are interconnected via a bus 1204. An Input / Output (I / O) interface 1205 is also connected to the bus 1204.

[0205] The following components are connected to I / O interface 1205: an input section 1206 including a keyboard, mouse, etc.; an output section 1207 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1208 including a hard disk, etc.; and a communication section 1209 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to I / O interface 1205 as needed. Removable media 1211, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1210 as needed so that computer programs read from them can be installed into storage section 1208 as needed.

[0206] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1209, and / or installed from removable medium 1211. When the computer program is executed by central processing unit (CPU) 1201, it performs various functions defined in the system of this application.

[0207] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0208] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0209] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0210] In another aspect, this application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable storage medium carries computer-readable instructions that, when executed by a processor, implement the methods in any of the above embodiments.

[0211] According to one aspect of the embodiments of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the electronic device to perform the methods of any of the above embodiments.

[0212] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0213] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause an electronic device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the methods according to the embodiments of this application.

[0214] Other embodiments of this application will readily conceive of by those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. It should be understood that this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

[0215] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An information processing method, characterized in that, The method includes: Based on the information attributes of each new piece of information and the total number of information under each preset information category, the proportion of each new piece of information under each preset information category is determined, and the proportion of each new piece of information under each preset information category is used as the probability that the information category of each new piece of information belongs to each preset information category. Based on the probability that the information category of each newly added piece of information belongs to each of the preset information categories, a new category scoring matrix is ​​constructed; Singular value decomposition is performed on the existing rating matrix to obtain the information interest matrix, user interest matrix, and covariance matrix; the existing rating matrix includes the probability that each user selects existing information under each preset information category; the existing information and the newly added information are products, songs, or travel service products; A new information interest matrix is ​​obtained based on the information interest matrix and the newly added category rating matrix; Based on the covariance matrix, the user interest matrix, and the new information interest matrix, an information rating matrix is ​​obtained. Obtain the rating thresholds for target information for each of multiple users, wherein the target information is at least one of the newly added information, and the information category of the newly added information belongs to at least one of multiple preset information categories; The information ratings for each of the multiple users corresponding to the target information are determined from the information rating matrix and used as the target ratings for each user. Each element in the information rating matrix represents a user's rating for a preset information category. If the user's target score is higher than the user's score threshold, the user is designated as the target user, and the target information is sent to the target user.

2. The method according to claim 1, characterized in that, The step of obtaining a new information interest matrix based on the information interest matrix and the newly added category rating matrix includes: Calculate the product of the information interest matrix and the inverse of the newly added category rating matrix, and use it as the conditional newly added category rating matrix; The product of the newly added category rating matrix and the newly added category rating matrix under the given conditions is calculated as the new information interest matrix.

3. The method according to claim 1, characterized in that, The step of obtaining the information rating matrix based on the covariance matrix, the user interest matrix, and the new information interest matrix includes: The product of the covariance matrix, the transpose of the user interest matrix, and the new information interest matrix is ​​calculated and used as the information rating matrix.

4. The method according to claim 1, characterized in that, The step of obtaining the rating thresholds for target information from multiple users includes: Determine the information category of the target information; The average rating of the information category corresponding to the target information is determined from the average rating matrix and used as the rating threshold. The average rating matrix includes the average rating of each of the multiple users for each preset information category. The average rating in the i-th row and j-th column of the average rating matrix represents the attention of the i-th user to the information in the j-th preset information category. i is any integer value between 1 and N, j is any integer value between 1 and M, N is the total number of the multiple users, and M is the total number of the multiple preset information categories.

5. The method according to claim 4, characterized in that, The method for obtaining the average rating matrix includes: Obtain target sample features for each of multiple users regarding the target sample information, wherein the target sample information is determined from the sample information and does not include the newly added information; The target sample features are processed by an information scoring model to obtain the predicted scores of each user for each target sample information. The predicted score of each user for each target sample information represents the probability that the target sample information is selected by the user. The average rating matrix is ​​obtained based on the predicted ratings given by each of the multiple users for each of the target sample information.

6. The method according to claim 5, characterized in that, The step of obtaining the average rating matrix based on the predicted ratings of the multiple users for each of the target sample information includes: If the user's predicted score for the target sample information reaches the user's target score threshold for the target sample information, it is determined that the target sample information has been selected by the user. The ratio of the number of target sample information selected by the nth user under the mth preset information category to the total number of all target sample information under the mth preset information category is used as the probability that the target sample information under the mth preset information category is selected by the nth user. m takes integer values ​​from 1 to M, n takes integer values ​​from 1 to N, M is the total number of the multiple preset information categories, and N is the total number of the multiple users. The average rating matrix is ​​constructed based on the probability that each user selects the target sample information under each preset information category.

7. The method according to claim 5, characterized in that, The training method for the information scoring model includes: Obtain sample category features for the multiple preset information categories, wherein the sample category features do not include the features of the newly added information; The sample category features are divided into sparse features and dense features; The sparse features are processed by the deep network layers in the initial model to obtain the first feature; The dense features are processed through the linear network layer in the initial model to obtain the second feature; The first feature and the second feature are fused to obtain the fused feature; The initial model is trained based on the fusion features to obtain the information scoring model.

8. The method according to claim 1, characterized in that, The methods for obtaining the existing rating matrix include: Obtain the selection results of each of the multiple users for the existing information, and the selection result of each user includes whether the existing information is selected by the user or not selected by the user; Calculate the ratio of the number of existing information selected by the qth user under the p-th preset information category to the total number of all existing information under the p-th preset information category, and use this as the probability that existing information under the p-th preset information category is selected by the qth user. p takes values ​​from 1 to M, q takes values ​​from 1 to N, M is the total number of the multiple preset information categories, and N is the total number of the multiple users. The existing rating matrix is ​​constructed based on the probability that each user selects existing information under each preset information category.

9. An information processing device, characterized in that, The device includes: The newly added category scoring matrix acquisition module is used to determine the proportion of each new piece of information in each preset information category based on the information attributes of each new piece of information and the total number of information in each preset information category, and to use the proportion of each new piece of information in each preset information category as the probability that the information category of each new piece of information belongs to each preset information category; and to construct a new category scoring matrix based on the probability that the information category of each new piece of information belongs to each preset information category. The information rating matrix acquisition module is used to perform singular value decomposition on an existing rating matrix to obtain an information interest matrix, a user interest matrix, and a covariance matrix. The existing rating matrix includes the probability that existing information under each preset information category is selected by each user. A new information interest matrix is ​​obtained based on the information interest matrix and the newly added category rating matrix. The information rating matrix is ​​obtained based on the covariance matrix, the user interest matrix, and the new information interest matrix. The existing information and the newly added information are products, songs, or travel services. The acquisition module is used to acquire the rating thresholds of multiple users for target information, wherein the target information is at least one of the newly added information, and the information category of the newly added information belongs to at least one of multiple preset information categories. The determining module is used to determine the scores of the information categories corresponding to the target information for each of the plurality of users from the information scoring matrix, and use them as the target scores for each of the users. In the information scoring matrix, each element represents a user's score for a preset information category. The sending module is used to designate the user as the target user and send the target information to the target user if the user's target score is higher than the user's score threshold.

10. The apparatus according to claim 9, characterized in that, The information scoring matrix acquisition module is also used for: Calculate the product of the information interest matrix and the inverse of the newly added category rating matrix, and use it as the conditional newly added category rating matrix; The product of the newly added category rating matrix and the newly added category rating matrix under the given conditions is calculated as the new information interest matrix.

11. The apparatus according to claim 9, characterized in that, The information scoring matrix acquisition module is also used for: The product of the covariance matrix, the transpose of the user interest matrix, and the new information interest matrix is ​​calculated and used as the information rating matrix.

12. The apparatus according to claim 9, characterized in that, The acquisition module is also used for: Determine the information category of the target information; The average rating of the information category corresponding to the target information is determined from the average rating matrix and used as the rating threshold. The average rating matrix includes the average rating of each of the multiple users for each preset information category. The average rating in the i-th row and j-th column of the average rating matrix represents the attention of the i-th user to the information in the j-th preset information category. i is any integer value between 1 and N, j is any integer value between 1 and M, N is the total number of the multiple users, and M is the total number of the multiple preset information categories.

13. The apparatus according to claim 12, characterized in that, The device further includes an average score matrix acquisition module, used for: Obtain target sample features for each of multiple users regarding the target sample information, wherein the target sample information is determined from the sample information and does not include the newly added information; The target sample features are processed by an information scoring model to obtain the predicted scores of each user for each target sample information. The predicted score of each user for each target sample information represents the probability that the target sample information is selected by the user. The average rating matrix is ​​obtained based on the predicted ratings given by each of the multiple users for each of the target sample information.

14. The apparatus according to claim 13, characterized in that, The average rating matrix acquisition module is also used for: If the user's predicted score for the target sample information reaches the user's target score threshold for the target sample information, it is determined that the target sample information has been selected by the user. The ratio of the number of target sample information selected by the nth user under the mth preset information category to the total number of all target sample information under the mth preset information category is used as the probability that the target sample information under the mth preset information category is selected by the nth user. m takes integer values ​​from 1 to M, n takes integer values ​​from 1 to N, M is the total number of the multiple preset information categories, and N is the total number of the multiple users. The average rating matrix is ​​constructed based on the probability that each user selects the target sample information under each preset information category.

15. The apparatus according to claim 13, characterized in that, The device further includes a model training module for: Obtain sample category features for the multiple preset information categories, wherein the sample category features do not include the features of the newly added information; The sample category features are divided into sparse features and dense features; The sparse features are processed by the deep network layers in the initial model to obtain the first feature; The dense features are processed through the linear network layer in the initial model to obtain the second feature; The first feature and the second feature are fused to obtain the fused feature; The initial model is trained based on the fusion features to obtain the information scoring model.

16. An electronic device, characterized in that, include: One or more processors; Memory; One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs being configured to perform the method as described in any one of claims 1-8.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1-8.

18. A computer program product, characterized in that, It includes computer instructions, which, when executed by a processor, implement the method as described in any one of claims 1 to 8.

Images