CN112163165B

CN112163165B - Information recommendation method, device, equipment and computer readable storage medium

Info

Publication number: CN112163165B
Application number: CN202011131347.1A
Authority: CN
Inventors: 张晗; 马连洋; 衡阵
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-10-21
Filing date: 2020-10-21
Publication date: 2024-05-17
Anticipated expiration: 2040-10-21
Also published as: CN112163165A

Abstract

The embodiment of the application provides an information recommendation method, device and equipment and a computer readable storage medium, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring at least one first type of feature and at least one second type of feature of each piece of information to be recommended in an information set to be recommended, and performing feature cross processing on the at least one first type of feature to obtain a first prediction score; performing feature fusion processing on the at least one second type of features to obtain a second prediction score; performing prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended; and recommending at least one piece of information to be recommended in the information set to be recommended to a target object according to the tendency score of each piece of information to be recommended. According to the embodiment of the application, the popularity of the information to be recommended on the specific crowd can be more accurately depicted, and the information recommendation effect on the specific crowd is improved.

Description

Information recommendation method, device, equipment and computer readable storage medium

Technical Field

The embodiment of the application relates to the technical field of Internet, and relates to an information recommendation method, an information recommendation device and a computer readable storage medium.

Background

In the information flow recommendation scene, user information and content characteristics are indispensable to a recommendation system, the existing content characteristics are mostly defined based on prior information of the content, such as attributing videos to sports, movies and the like according to the video content, and are not linked with behaviors of users, and for the user side, the recommendation system can utilize historical consumption behaviors of the users except basic information of the users to construct user interest point characteristics corresponding to the content characteristics. And when recommending, predicting clicking behaviors by utilizing the characteristics of the user and the video content. However, the recommendation thought has a large problem in the recommendation of the new user, and because the new user has no history consumption behavior record, the recommendation system cannot acquire the interest point characteristics of the user, and only the user basic information can be used for recommendation, so that the recommendation system cannot accurately predict the content of interest of the new user. Meanwhile, the recommendation of the new user is very important, and the retention of the new user determines key indexes of the whole recommended product such as the scale of the user, so that the cold start problem of the new user is an important difficulty frequently faced by the existing recommendation system.

Aiming at the cold start problem of the new user, the technical scheme in the related technology can only use the basic attribute of the user to recommend. Recommending local news content based on the regional information of the user; or based on the information of the gender, age and the like of the user, counting the consumption content of the crowd under the specific gender and age, and recommending the high consumption content.

The solutions in the related art are only applicable to contents that have been consumed, statistics cannot be made for newly produced contents, and the recommendation effect for a specific crowd may be poor due to the complexity of the recommendation system.

Disclosure of Invention

The embodiment of the application provides an information recommendation method, device and equipment and a computer readable storage medium, and relates to the technical field of artificial intelligence. The feature cross processing and the feature fusion processing are respectively carried out on at least one first type feature and at least one second type feature of the information to be recommended, and the obtained first prediction score and second prediction score are subjected to prediction result transformation processing, so that the tendency score of each piece of information to be recommended is obtained, and the information recommendation is carried out according to the tendency score.

The technical scheme of the embodiment of the application is realized as follows:

The embodiment of the application provides an information recommendation method, which comprises the following steps:

Acquiring at least one first type feature and at least one second type feature of each piece of information to be recommended in the information set to be recommended; performing feature cross processing on the at least one first type of features to obtain a first prediction score; performing feature fusion processing on the at least one second type of features to obtain a second prediction score; performing prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended; and recommending at least one piece of information to be recommended in the information set to be recommended to a target object according to the tendency score of each piece of information to be recommended.

The embodiment of the application provides an information recommendation device, which comprises:

The acquisition module is used for acquiring at least one first type of characteristics and at least one second type of characteristics of each piece of information to be recommended in the information set to be recommended; the feature cross processing module is used for performing feature cross processing on the at least one first type of features to obtain a first prediction score; the feature fusion processing module is used for carrying out feature fusion processing on the at least one second type of features to obtain a second prediction score; the transformation processing module is used for carrying out prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended; and the recommending module is used for recommending at least one piece of information to be recommended in the information set to be recommended to a target object according to the tendency score of each piece of information to be recommended.

Embodiments of the present application provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium; the processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor is configured to execute the computer instructions to implement the information recommendation method.

An embodiment of the present application provides an information recommendation apparatus, including: a memory for storing executable instructions; and the processor is used for realizing the information recommendation method when executing the executable instructions stored in the memory.

The embodiment of the application provides a computer readable storage medium, which stores executable instructions for realizing the information recommendation method when a processor executes the executable instructions.

The embodiment of the application has the following beneficial effects: and carrying out feature cross processing and feature fusion processing on at least one first type feature and at least one second type feature respectively aiming at each piece of information to be recommended in the information to be recommended set to correspondingly obtain a first prediction score and a second prediction score, and carrying out prediction result conversion processing on the first prediction score and the second prediction score to obtain a tendency score of each piece of information to be recommended, so that at least one piece of information to be recommended in the information to be recommended set is recommended according to the tendency score of each piece of information to be recommended, namely the information to be recommended is recommended to a target object belonging to a specific crowd. Therefore, the final tendency score is determined by processing the first type of features and the second type of features based on the information to be recommended, so that the complexity and the diversity of the features of the information to be recommended are fully considered, the popularity of the information to be recommended on the specific crowd can be more accurately depicted, and the information recommendation effect on the specific crowd is improved.

Drawings

FIG. 1 is a schematic diagram of an alternative architecture of an information recommendation system according to an embodiment of the present application;

Fig. 2 is a schematic structural diagram of a server according to an embodiment of the present application;

FIG. 3 is a schematic flow chart of an alternative information recommendation method according to an embodiment of the present application;

FIG. 4 is a schematic flow chart of an alternative information recommendation method according to an embodiment of the present application;

FIG. 5 is a schematic flow chart of an alternative information recommendation method according to an embodiment of the present application;

FIG. 6 is a schematic flow chart of an alternative information recommendation method according to an embodiment of the present application;

FIG. 7 is a schematic flow chart of an alternative information recommendation method according to an embodiment of the present application;

FIG. 8 is a schematic flow chart of an alternative information recommendation method according to an embodiment of the present application;

Fig. 9A is a block diagram of a tendency recognition network provided by an embodiment of the present application;

FIG. 9B is a flowchart illustrating an alternative method for training a trend identification network according to an embodiment of the present application;

fig. 10 is a network configuration diagram of an overall algorithm in an embodiment of the present application.

Detailed Description

The present application will be further described in detail with reference to the accompanying drawings, for the purpose of making the objects, technical solutions and advantages of the present application more apparent, and the described embodiments should not be construed as limiting the present application, and all other embodiments obtained by those skilled in the art without making any inventive effort are within the scope of the present application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which embodiments of this application belong. The terminology used in the embodiments of the application is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

In order to better understand the network structure searching method provided in the embodiment of the present application, first, a description is given of a network structure searching method in the related art:

In the related art, for the cold start problem of a new user, only the basic attribute of the user can be used for recommendation. However, this solution is only applicable to content that has already been consumed, and no statistics can be made for newly produced content. Therefore, the application provides the crowd tendency label of the content aiming at the pain point, and only the characteristics of the content are utilized to predict the popularity of the content to specific crowds. Before the content is recommended, the crowd identification popular with the content can be provided for the recommendation system, and the recommendation system is guided to recommend new users under specific crowd, namely the cold start problem of the new users is relieved.

In addition, the complexity of the recommendation system may result in poor recommendation effect for a certain crowd, for example, in the case that female users occupy less space, the recommendation system may not be able to learn the clicking behavior of the user crowd better. Therefore, in the scenario of optimizing the recommendation effect for the specific crowd, the prediction of crowd tendency is also necessary, for example, the operation can be assisted to perform content screening of the tendency of the specific crowd.

Based on the demand and problem analysis of the related art, the application focuses on solving the following problems: the information recommendation method is provided, and firstly, the group popularity of content (namely information to be recommended) is defined based on consumption indexes to be optimized of a recommendation system, such as click rate, page View (PV), duration, sharing and the like. If optimizing is performed based on the click rate, dividing the video content consumed by the group into two categories, namely popular and unpopular according to the click rate; secondly, constructing a wide & deep classification model by using the constructed classification data and multi-mode features such as titles, classifications, account numbers, labels, covers, video frames and the like of the video; and finally, predicting the content by using the trained model, and using the tendency score output by the last layer of the classification network as the tendency score of the content on the crowd.

The information recommendation method provided by the embodiment of the application comprises the steps of firstly, acquiring at least one first type of characteristics and at least one second type of characteristics of each piece of information to be recommended in an information set to be recommended; then, performing feature cross processing on at least one first type of features to obtain a first prediction score; performing feature fusion processing on at least one second type of features to obtain a second prediction score; performing prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended; and finally, recommending at least one piece of information to be recommended in the information set to be recommended to the target object according to the tendency score of each piece of information to be recommended. Therefore, the final tendency score is determined by processing the first type of features and the second type of features based on the information to be recommended, so that the complexity and the diversity of the features of the information to be recommended are fully considered, the popularity of the information to be recommended on the specific crowd can be more accurately depicted, and the information recommendation effect on the specific crowd is improved.

An exemplary application of the information recommendation device according to the embodiment of the present application is described below, in one implementation manner, the information recommendation device provided by the embodiment of the present application may be implemented as any terminal such as a notebook computer, a tablet computer, a desktop computer, a mobile device (for example, a mobile phone, a portable music player, a personal digital assistant, a dedicated messaging device, a portable game device), an intelligent robot, and in another implementation manner, the information recommendation device provided by the embodiment of the present application may be implemented as a server. In the following, an exemplary application when information recommendation is implemented as a server will be described.

Referring to fig. 1, fig. 1 is a schematic diagram of an alternative architecture of an information recommendation system 10 according to an embodiment of the present application. In order to recommend the generated video to a user or a new user interested in the video, the information recommendation system 10 provided in the embodiment of the present application includes a terminal 100, a network 200 and a server 300, wherein the terminal 100 is provided with a video recommendation application, the user can register on the video recommendation application to become the new user, after successful registration, the video recommendation application actively recommends the video to the user, and the recommended video is the video which may be interested in the user, at this time, the terminal 100 can send a registration completion message to the server 300 through the network 200 to inform the current user to be the new user, and the server 300 acquires at least one first type feature and at least one second type feature of each video to be recommended in the video set to be recommended; then, performing feature cross processing on at least one first type of features to obtain a first prediction score; performing feature fusion processing on at least one second type of features to obtain a second prediction score; performing prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the video to be recommended; and determining at least one video to be recommended in the video to be recommended set according to the tendency score of each video to be recommended, and recommending the videos to be recommended to the terminal 100.

The information recommendation method provided by the embodiment of the application also relates to the technical field of artificial intelligence, and can be at least realized through machine learning and natural language processing technologies in the artificial intelligence technology. The machine learning (M L, MACHINE LEARNING) is a multi-domain interdisciplinary, and relates to multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. It is specially studied how a computer simulates or implements learning behavior of a human to acquire new knowledge or skills, and reorganizes existing knowledge structures to continuously improve own performance. Machine learning is the core of artificial intelligence, a fundamental approach to letting computers have intelligence, which is applied throughout various areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, confidence networks, reinforcement learning, transfer learning, induction learning, teaching learning, and the like. In the embodiment of the application, the response to the network structure search request is realized through a machine learning technology, so that the target network structure is automatically searched, and the training and model optimization of the controller and the score model are realized. Natural language processing (N LP, nature Language processing) is an important direction in the fields of computer science and artificial intelligence, and it is studying various theories and methods that enable efficient communication between humans and computers in natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Thus, the research in this field will involve natural language, i.e. language that people use daily, so it has a close relationship with the research in linguistics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic questions and answers, knowledge graph techniques, and the like.

Fig. 2 is a schematic structural diagram of a server 300 according to an embodiment of the present application, where the server 300 shown in fig. 2 includes: at least one processor 310, a memory 350, at least one network interface 320, and a user interface 330. The various components in server 300 are coupled together by bus system 340. It is understood that the bus system 340 is used to enable connected communications between these components. The bus system 340 includes a power bus, a control bus, and a status signal bus in addition to the data bus. But for clarity of illustration the various buses are labeled in fig. 2 as bus system 340.

The Processor 310 may be an integrated circuit chip with signal processing capabilities such as a general purpose Processor, such as a microprocessor or any conventional Processor, a digital signal Processor (DSP, digital Signal Processor), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like.

The user interface 330 includes one or more output devices 331 that enable presentation of media content, including one or more speakers and/or one or more visual displays. The user interface 330 also includes one or more input devices 332, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

Memory 350 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid state memory, hard drives, optical drives, and the like. Memory 350 optionally includes one or more storage devices physically located remote from processor 310. Memory 350 includes volatile memory or nonvolatile memory, and may also include both volatile and nonvolatile memory. The non-volatile Memory may be a Read Only Memory (ROM) and the volatile Memory may be a random access Memory (RAM, random Access Memory). The memory 350 described in embodiments of the present application is intended to comprise any suitable type of memory. In some embodiments, memory 350 is capable of storing data to support various operations, examples of which include programs, modules and data structures, or subsets or supersets thereof, as exemplified below.

The operating system 351 including system programs for handling various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and handling hardware-based tasks;

Network communication module 352 for reaching other computing devices via one or more (wired or wireless) network interfaces 320, exemplary network interfaces 320 include: bluetooth, wireless compatibility authentication (WiFi), and universal serial bus (USB, universal Serial Bus), etc.;

An input processing module 353 for detecting one or more user inputs or interactions from one of the one or more input devices 332 and translating the detected inputs or interactions.

In some embodiments, the apparatus provided in the embodiments of the present application may be implemented in software, and fig. 2 shows an information recommendation device 354 stored in a memory 350, where the information recommendation device 354 may be an information recommendation device in a server 300, and may be software in the form of a program and a plug-in, and includes the following software modules: the acquisition module 3541, feature intersection processing module 3542, feature fusion processing module 3543, transformation processing module 3544, and recommendation module 3545 are logical, and thus can be arbitrarily combined or further split depending on the implemented functionality. The functions of the respective modules will be described hereinafter.

In other embodiments, the apparatus provided by the embodiments of the present application may be implemented in hardware, and by way of example, the apparatus provided by the embodiments of the present application may be a processor in the form of a hardware decoding processor that is programmed to perform the information recommendation method provided by the embodiments of the present application, for example, the processor in the form of a hardware decoding processor may employ one or more Application Specific Integrated Circuits (ASICs), DSPs, programmable logic devices (PLDs, programmable Logic De vice), complex Programmable logic devices (CPLDs, complex Programmable Logic Device), field-Programmable gate arrays (FPGAs), or other electronic components.

The information recommendation method provided by the embodiment of the present application will be described below in connection with exemplary applications and implementations of the server 300 provided by the embodiment of the present application. Referring to fig. 3, fig. 3 is a schematic flowchart of an alternative information recommendation method according to an embodiment of the present application, and will be described with reference to the steps shown in fig. 3.

Step S301, at least one first type feature and at least one second type feature of each information to be recommended in the information to be recommended set are obtained.

The information to be recommended in the information to be recommended set may have a certain correlation, for example, each information to be recommended in the information to be recommended set has an interactive relationship with a specific target object, that is, each information to be recommended in the information to be recommended set has interactive data with any user group, and a user in the user group clicks or browses each information to be recommended in the information to be recommended set, or each information to be recommended in the information to be recommended set has been recommended to a user in the user group.

In the embodiment of the application, when information recommendation is performed, at least one piece of information with recommendation information in the information set to be recommended is recommended to the target user, so that each piece of information to be recommended in the information set to be recommended needs to be analyzed to determine whether the information to be recommended can be recommended or not.

In the embodiment of the application, when each piece of information to be recommended is analyzed, the information to be recommended is based on the first type of characteristics and the second type of characteristics of the information to be recommended. For example, the first class of features may be low-order features, and the second class of features may be high-order features, where the low-order features are features calculated or determined from fewer original features in the information to be recommended, and the high-order features are features calculated or determined from more original features in the information to be recommended.

For example, when the information to be recommended is a video, the first type of features may be low-order features that may be determined by a first number of original features of the video, such as a category of the video, a duration of the video, or a quality of the video, and the second type of features may be high-order features that may be determined by a second number of original features of the video, such as a title, a cover, or a label of the video.

In some embodiments, the number of original features of the information to be recommended related to the first type of features may be smaller than the number of original features of the information to be recommended related to the second type of features, for example, the first type of features needs to be determined by a first number of original features in the information to be recommended, and the second type of features needs to be determined by a second number of original features in the information to be recommended, where the first number may be smaller than the second number. Of course, the number of the original features of the information to be recommended related to the first type of features may be greater than or equal to the number of the original features of the information to be recommended related to the second type of features, and may be specifically determined according to the input information of the network structure in the information recommendation model, which is not particularly limited in the embodiment of the present application.

Step S302, performing feature cross processing on at least one first type of feature to obtain a first prediction score.

Here, after the first type of feature is acquired, a feature cross process is performed on the first type of feature, where the feature cross process may be implemented by: firstly, carrying out feature embedding processing on each first type of feature to obtain embedded vectors, then, carrying out linear transformation on the embedded vectors corresponding to different first types of features, for example, carrying out calculation such as vector multiplication or vector summation, so as to realize feature cross processing on the embedded vectors corresponding to a plurality of first types of features, and obtaining a final output result, wherein the output result is a one-dimensional vector, and the output one-dimensional vector is the first prediction component.

And step S303, performing feature fusion processing on at least one second type of features to obtain a second prediction score.

Here, after the second type of feature is acquired, feature fusion processing is performed on the second type of feature, where the feature fusion processing may be implemented by: since the second class of features may include at least a text class feature and an image class feature, processing may be performed for the text class feature and the image class feature, respectively. Firstly, word embedding processing is carried out on text type features in the second type features to obtain word embedding vectors of each word, and then the average value of all word embedding vectors is calculated, so that the encoding process of the text type features in the second type features is realized; for the image type features in the second type features, a specific feature extraction network can be directly adopted for feature extraction to obtain vectors after image type feature coding; and then, respectively carrying out weighting treatment and linear transformation on the coded vectors of the text type features and the image type features, then carrying out splicing and fusion, and finally carrying out linear transformation to reduce the dimension of the spliced vectors to one dimension to obtain a one-dimension vector, wherein the one-dimension vector is the second prediction component.

In some embodiments, when the text class feature included in the second class feature is composed of a plurality of words, for example, when the text class feature is a sentence or an article, since the sentence or the article is a text composed of a plurality of words, before the feature fusion process is performed, it is first necessary to perform a word segmentation process on the sentence or the article, and then perform the above word embedding process.

In the embodiment of the application, the text type features can comprise features composed of a plurality of words, such as a title, a text and the like, and can also comprise tag features composed of one word. The word segmentation processing can be carried out only on the features such as the title, the text and the like formed by a plurality of words, and then word embedding processing is carried out on each word obtained after the word segmentation processing to obtain word embedding vectors; since the tag is usually a single word, word embedding processing can be directly performed to obtain a word embedding vector without word segmentation processing on the tag features.

And step S304, performing prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended.

Here, the first prediction score and the second prediction score are integrated, and the first prediction score and the second prediction score are summed and transformed to obtain the tendency score of the information to be recommended.

In step S305, at least one piece of information to be recommended in the information set to be recommended is recommended to the target object according to the tendency score of each piece of information to be recommended.

In the embodiment of the application, the information to be recommended with the tendency score being greater than the threshold value can be recommended to the target object, for example, the value of the tendency score can be any value between 0 and 1, and the threshold value can be 0.5; or ordering the information to be recommended in the information set to be recommended according to the order of the tendency scores from large to small to form an information sequence to be recommended, and then recommending the specific quantity of information to be recommended, which is positioned before the information sequence to be recommended, to the target object according to the order of the information sequence to be recommended.

The target object may be an object belonging to a particular group, for example, the target object may be a new user, or may be a female group user.

According to the information recommending method provided by the embodiment of the application, for each piece of information to be recommended in the information to be recommended set, at least one first type of feature and at least one second type of feature are subjected to feature cross processing and feature fusion processing respectively to correspondingly obtain a first prediction score and a second prediction score, and the first prediction score and the second prediction score are subjected to prediction result conversion processing to obtain a tendency score of each piece of information to be recommended, so that at least one piece of information to be recommended in the information to be recommended set is recommended according to the tendency score of each piece of information to be recommended, namely the information to be recommended is recommended to a target object belonging to a specific crowd. Therefore, the final tendency score is determined by processing the first type of features and the second type of features based on the information to be recommended, so that the complexity and the diversity of the features of the information to be recommended are fully considered, the popularity of the information to be recommended on the specific crowd can be more accurately depicted, and the information recommendation effect on the specific crowd is improved.

In some embodiments, the method of the embodiment of the present application may also be applied to an initial recommendation process after a new user completes registration, where the information recommendation system includes a terminal and a server, the user may register on the terminal to become a new user of any application, and after the user has completed registration, the application performs information recommendation to the user to further obtain a consumption log of the user, where the consumption log includes but is not limited to: click frequency, viewing duration, number of exposures, etc.

Fig. 4 is a schematic flow chart of an alternative information recommendation method provided in an embodiment of the present application, as shown in fig. 4, the method includes the following steps:

in step S401, the terminal receives a registration request of a user.

In step S402, after the user completes registration on the client of the specific user on the terminal, the terminal sends an information recommendation request to the server, where the information recommendation request includes the basic information of the user.

Here, the basic information of the user is information input in the user registration process, for example, information of contact, sex, age, current location, and the like.

In step S403, the server obtains at least one first type feature and at least one second type feature of each piece of information to be recommended in the set of information to be recommended.

In step S404, the server performs feature cross processing on at least one first class feature to obtain a first prediction score.

And step S405, the server performs feature fusion processing on at least one second type of features to obtain a second prediction score.

In step S406, the server performs a prediction result transformation process on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended.

In step S407, the server selects at least one piece of information to be recommended from the set of information to be recommended according to the tendency score of each piece of information to be recommended.

It should be noted that, the steps S403 to S403 are the same as the steps S301 to S305, and the embodiments of the present application are not described in detail.

In step S408, the server recommends the selected at least one piece of information to be recommended to the terminal.

According to the information recommending method provided by the embodiment of the application, after the new user is registered, because the new user does not have interactive information in the application, the interest point information of the user cannot be determined according to the use habit of the new user, so that the information to be recommended with higher tendency score is selected from the information to be recommended set by adopting the method provided by the embodiment of the application, and is the information possibly interested by the new user, and the information is recommended to the new user, so that the accuracy of primary recommendation can be improved, and the recommending effect can be improved.

In some embodiments, before information recommendation is performed, a set of information to be recommended needs to be determined first, so that at least one piece of information to be recommended may be determined from the set of information to be recommended during information recommendation, based on fig. 3, fig. 5 is an optional flowchart of an information recommendation method provided in an embodiment of the present application, as shown in fig. 5, before step S301, the method further includes the following steps:

In step S501, attribute information of the target object is determined. The attribute information of the target object may be information of the type or characteristics of the target object, for example, the attribute information of the target object may be that the target object is a female user, the age of the target object is between 20 and 30 years, or the like.

Step S502, an object set formed by the objects with the attribute information and a recommendation information set formed by recommendation information with interaction information between the objects in the object set are obtained.

Here, each object in the object set has the attribute information, for example, the objects in the object set may all be female users; each piece of recommended information in the recommended information set has interaction information with any object in the object set, and the interaction information can be information corresponding to any interaction operation such as clicking, browsing, purchasing, sharing and the like, for example, information purchased by a female user in the female user set can be used as recommended information to form the recommended information set.

In step S503, for each piece of recommended information in the set of recommended information, interaction data between each object in the set of objects and the recommended information is obtained.

Here, taking the interaction data as the click data as an example, the number of clicks of each female user in the female user set on each recommended information is obtained.

Step S504, aggregating the interactive data to obtain the interactive probability of the object in the object set for each recommended information.

Here, after the click times of all the female users are summed, the average is performed in the female user set, so as to obtain the click probability of the female users of the female user set for each piece of recommended information.

In step S505, at least one piece of recommended information is selected from the recommended information set according to the interaction probability, so as to form the recommended information set.

Here, the recommended information with the interaction probability greater than the probability threshold is selected as the information to be recommended, or the recommended information with the preset number is selected as the information to be recommended according to the order of the interaction probability from high to low.

In some embodiments, the interactive data includes at least exposure times, click times, and browsing durations for the recommended information; based on fig. 5, fig. 6 is a schematic flowchart of an alternative information recommendation method according to an embodiment of the present application, as shown in fig. 6, step S504 may be implemented by:

Step S601, aggregating the exposure times, the clicking times and the browsing time length to obtain an average clicking rate and an average browsing time length for each piece of recommended information.

Here, the average click rate is an average value of the total number of clicks corresponding to the objects in the object set over the entire object set, that is, a value obtained by dividing the total number of clicks by the number of objects in the object set; the average browsing duration is an average value of the total browsing duration corresponding to the objects in the object set over the whole object set, that is, a value obtained by dividing the total browsing duration by the number of the objects in the object set.

Step S602, determining any one of the average click rate and the average browsing duration as the interaction probability of the recommended information.

With continued reference to fig. 6, in some embodiments, the method further comprises the steps of:

step S603, determining the total average click rate corresponding to all the objects in the object set according to the average click rate of each piece of recommendation information in the recommendation information set.

Here, the total average click rate refers to an average value of the average click rate of each recommended information in the entire recommended information set, that is, a value obtained by dividing the sum of average click rates corresponding to all objects by the number of recommended information in the recommended information set.

Step S604, determining the total average browsing duration corresponding to all the objects in the object set according to the average browsing duration of each piece of recommended information in the recommended information set.

Here, the total average browsing duration refers to an average value of the average browsing duration of each recommended information in the whole recommended information set, that is, a value obtained by dividing the sum of average browsing durations corresponding to all objects by the number of recommended information in the recommended information set.

With continued reference to fig. 6, in some embodiments, step S505 may be worth any one of the following implementations: mode one: in step S605, in the recommendation information set, recommendation information with an average click rate greater than the total average click rate is selected as information to be recommended.

Square picking up: in step S606, the recommended information with the average browsing time period longer than the total average browsing time period is selected from the recommended information set as the information to be recommended.

In some embodiments, the first class of features includes at least one of an information class feature, an information length feature, and an information quality feature; when the first type of features only comprise information type features, performing feature cross processing based on the information type features to obtain a first prediction score; when the first type of features only comprise information length features, performing feature cross processing based on the information length features to obtain a first prediction score; and when the first type of features only comprise the information quality features, performing feature intersection processing based on the information quality features to obtain a first prediction score. Of course, the first type of features may also include an information category feature, an information length feature, and an information quality feature, and then feature cross processing may be performed based on the information category feature, the information length feature, and the information quality feature to obtain the first prediction score. The following description will take an example in which the first type of features includes an information category feature, an information length feature, and an information quality feature.

Based on fig. 3, fig. 7 is a schematic flowchart of an alternative information recommendation method according to an embodiment of the present application, as shown in fig. 7, step S302 may be implemented by:

and step S701, respectively carrying out coding processing on the first class feature and the second class feature in the information class features to correspondingly obtain a first class feature vector and a second class feature vector.

Here, the information category feature refers to that the information to be recommended is attributed to a certain category according to the content of the information to be recommended, where the category may be a primary category and a secondary category, respectively, the primary category is a parent category of the secondary category, and each information to be recommended may have not only one primary category but also one secondary category.

When the information category characteristics of the information to be recommended are acquired, namely the primary category characteristics and the secondary category characteristics of the information to be recommended are acquired, the primary category characteristics and the secondary category characteristics are respectively subjected to coding processing, and a primary category characteristic vector and a secondary category characteristic vector are obtained. In some embodiments, the encoding process herein may employ a single-hot encoding approach to encoding.

Step S702, discretizing and encoding the information length feature in sequence to obtain an information length feature vector.

Because the information length feature may be a continuous feature, the continuous information length feature may be first discretized, for example, discretized by using an equal frequency bin to obtain a plurality of discretized levels, and then the discretized plurality of features may be encoded to obtain an information length feature vector.

In step S703, the information quality feature is encoded to obtain an information quality feature vector.

In some embodiments, step S703 may also be implemented by:

in step S7031, a pre-scoring value for information to be recommended is acquired.

Step S7032, the pre-scoring value is encoded, and an information quality feature vector is obtained.

In the embodiment of the application, the information quality can be scored first to obtain the pre-scored value of the information to be recommended, namely the information to be recommended can be scored according to at least one priori standard such as the authenticity of the information, the sufficiency of the content, the precision and the chroma of the content and the like, then the scoring result is classified into a plurality of grades, and the plurality of grades are subjected to coding processing to obtain the information quality feature vector.

In step S704, a factorizer is used to perform cross processing on the first class feature vector, the second class feature vector, the information length feature vector and the information quality feature vector, so as to obtain a first prediction score.

Here, the factorizer is configured to perform linear prediction on the first class feature vector, the second class feature vector, the information length feature vector, and the information quality feature vector, and finally output a one-dimensional vector, which is a first prediction component.

It should be noted that, the step of determining the first prediction score may also be implemented by using a wide network in the wide & deep networks.

In some embodiments, the second type of features includes at least text features and image features; with continued reference to fig. 7, step S303 may be implemented by:

Step S705, determining at least one word according to the text feature.

Here, when the text feature is a title or a body, at least one word may be obtained by performing word segmentation processing on the title or the body; when the text feature is a label, the word corresponding to the label can be directly acquired.

Step S706, a word embedding vector for each word is acquired.

Step S707, determining the average value of the word embedding vectors as the encoding vector of the text feature.

In step S708, image feature extraction is performed on the image features to obtain image feature vectors.

In step S709, feature fusion processing is performed on the encoded vector and the image feature vector based on the attention mechanism, so as to obtain a second prediction score.

With continued reference to fig. 7, in some embodiments, step S304 may be implemented by:

in step S710, the first prediction score and the second prediction score are summed to obtain a prediction total score.

And step S711, performing prediction result transformation processing on the prediction total score by adopting a nonlinear activation function to obtain a tendency score of the information to be recommended.

Based on fig. 7, fig. 8 is a schematic flow chart of an alternative information recommendation method provided in an embodiment of the present application, as shown in fig. 8, step S709 may be implemented by:

in step S801, based on the attention mechanism, the encoding vector and the image feature vector are respectively subjected to linear transformation, so as to correspondingly obtain the encoding vector after linear transformation and the image feature vector after linear transformation.

In some embodiments, step S801 may also be implemented by:

In step S8011, a first attention weight of the encoding vector and a second attention weight of the image feature vector are determined. In step S8012, the encoding vector and the image feature vector are weighted by the first attention weight and the second attention weight, respectively. The weighting here refers to multiplying the first attention weight by the encoding vector and the second attention weight by the image feature vector.

In step S8013, the weighted encoding vector and the weighted image feature vector are respectively subjected to linear transformation, so as to correspondingly obtain the linear transformed encoding vector and the linear transformed image feature vector.

Here, FC linear transformation is performed on the weighted encoding vector and the weighted image feature vector by using a Fully Connected network (FC).

Step S802, splicing the encoding vector after linear transformation and the image characteristic vector after linear transformation to obtain a spliced vector.

And step 803, performing activation processing on the spliced vector to reduce the dimension of the spliced vector to one dimension, so as to obtain a second prediction score. Here, the activation processing may be performed on the spliced vector by using the relu function, and after the activation processing is performed on the relu function, a one-dimensional vector is connected, so that the dimension of the spliced vector after the activation processing is reduced to one dimension, and a second prediction score is obtained.

It should be noted that, the step of determining the second prediction score may also be implemented by using a deep network in the wide & deep network.

In some embodiments, a tendency recognition network may be trained in advance, and the step of determining the tendency score of the information to be recommended in any of the above embodiments may be implemented through the trained tendency recognition network. Based on this, the embodiment of the present application provides a trend identifying network and a training method of the trend identifying network, fig. 9A is a block diagram of the trend identifying network provided by the embodiment of the present application, as shown in fig. 9A, the trend identifying network 90 includes a feature cross network 901, a feature fusion network 902 and a transformation processing layer 903, where the feature cross network 901 is used for performing feature cross processing on a first type of features of input information to be recommended to obtain a first prediction score; the feature fusion network 902 is configured to perform feature fusion processing on the second type of features of the input information to be recommended to obtain a second prediction score; the transformation processing layer 903 is configured to perform a prediction result transformation process on the first prediction score output by the feature cross network 901 and the second prediction score output by the feature fusion network 902, so as to obtain a tendency score of the information to be recommended.

Fig. 9B is a schematic flowchart of an alternative training method of a trend identification network according to an embodiment of the present application, where, as shown in fig. 9B, the training method includes the following steps:

In step S901, a first type of sample feature of the sample information and a second type of sample feature of the sample information are input into the tendency recognition network.

Step S902, performing feature cross processing on the first type sample features through a feature cross network of the tendency identification network to obtain a first sample prediction score.

And step 903, performing feature fusion processing on the second type sample features through a feature fusion network of the tendency identification network to obtain a second sample prediction score.

In step S904, the prediction result transformation process is performed on the first sample prediction score and the second sample prediction score by the transformation processing layer of the tendency recognition network, so as to obtain a sample tendency score of the sample information.

In step S905, the sample tendency score is input into a predetermined loss model to obtain a loss result.

Here, the preset loss model is used for comparing the sample tendency score with a preset tendency score to obtain a loss result, where the preset tendency score may be a real tendency score preset by a user and corresponding to sample information, and the preset tendency score may take a value of 0 or 1.

In the embodiment of the application, the preset loss model comprises a loss function, the distance between the sample tendency score and the preset tendency score is calculated through the loss function, and the loss result is determined according to the distance. When the distance between the sample tendency score and the preset tendency score is larger, the difference between the output result of the model and the true value is larger, and further training is needed; when the distance between the sample tendency score and the preset tendency score is smaller, the smaller the difference between the training result of the model and the true value is, the smaller the difference between the output result of the model and the true value is, the result is closer to the true value, and further training of the model can be considered to be stopped.

Step S906, correcting parameters in the feature crossing network and the feature fusion network according to the loss result to obtain a trained tendency identification network.

Here, when the distance is smaller than the distance threshold, the loss result indicates that the current feature intersection network cannot perform accurate feature intersection processing on the first type of sample features, or indicates that the current feature fusion network cannot perform accurate feature fusion processing on the second type of sample features, so that parameters in at least one of the feature intersection network and the feature fusion network need to be corrected until the distance between the sample tendency score output by the tendency recognition network and the preset tendency score is smaller than the distance threshold, and training on the tendency recognition network is stopped.

According to the training method of the tendency recognition network, as the sample information is input into the tendency recognition network, the characteristic crossing processing, the characteristic fusion processing and the prediction result transformation processing are sequentially carried out on the sample information through the characteristic crossing network and the characteristic fusion network, so that the sample tendency score of the sample information is obtained, and the sample tendency score is input into the preset loss model, and the loss result is obtained. Therefore, parameters in the feature cross network and the feature fusion network can be corrected according to the loss result, and the obtained trained tendency recognition network can accurately predict the tendency score of the information to be recommended, so that accurate information recommendation is performed on the user.

In some embodiments, when model prediction is performed by using a tendency recognition network, each piece of information to be recommended may be sequentially input into the tendency recognition network, where the tendency recognition network may automatically output a tendency score corresponding to the information to be recommended, where the tendency score may be any value between 0 and 1, and when the tendency score is greater than 0.5, it indicates that the popularity of the corresponding information to be recommended is higher, and the information to be recommended may be recommended to the target user; when the tendency score is smaller than or equal to 0.5, the popularity of the corresponding information to be recommended is lower, and the information to be recommended is not recommended to the target user.

In the following, an exemplary application of the embodiment of the present application in a practical application scenario will be described.

The embodiment of the application provides an information recommendation method, which is a video crowd tendency prediction scheme based on multi-mode information, and aims to score the popularity of video content on specific crowds so as to improve the accuracy of personalized recommendation on the specific crowds. Firstly, modeling a concept of crowd tendency, and dividing contents into two types of likes and dislikes according to consumption indexes such as click rate, duration and the like by using consumption logs of the crowd, namely converting the popularity of the contents into two types of problems; secondly, constructing a wide & deep classification model to simulate the problems by utilizing multi-mode information of the video, such as titles, classifications, labels, covers, video frames and the like; finally, in the prediction, the output probability of the classification model is used as the popularity score (i.e., tendency score) of the content on the specified population. On the recommendation side, the score can be utilized to optimize the recommendation strategy for the specified crowd, such as increasing the recommendation weight of the high-popularity score video in the cold start stage of the new user, so as to further improve the consumption indexes such as clicking, duration, retention and the like of the specified crowd.

The trend data set (i.e. the information set to be recommended) used in the embodiment of the application is from a user consumption log under the scene of a main feeds of an information stream product, wherein the data format of the user consumption log is (t, uid _i,vid_v,exp_i,v,clk_i,v,playdur_i,v), wherein t is a reporting timestamp, uid _i is a user i, vid _v is a video v, exp _i,v is the exposure times of the video v to the user i, clk _i,v is the click times of the user i to the video v, and playdur _i,v is the watching duration of the video v by the user i. And (3) screening the logs according to the designated user group (assumed to be female groups), and only keeping records corresponding to female users. Defining a certain consumption interval, grouping according to vid _v, and aggregating consumption data such as exposure, clicking, duration and the like to obtain an aggregated data structure (vid _v,exp_v,clk_v,ctr_v,playdur_v), wherein the total exposure number of women of the video v isThe total number of clicks for video v is/>The average click rate of the video v is ctr _v＝clk_v/exp_v, and the average playing time length of the video v is/>U is the total set of female users. Further, the average click rate of the female population on the total content consumption is obtained as/>The average duration of the female population for total content consumption isThe female population corresponds to the object set, and the entire content corresponds to the recommendation information set.

In the embodiment of the application, a video set which is popular with females or not can be constructed according to the business target (assumed to be ctr) to be optimized. If it isThen video v is a positive sample, i.e., popular with women; and conversely, the negative sample is not welcomed by women. The data set (i.e. the set of information to be recommended) required to classify the female liabilities is thus obtained.

Because of the high impact of determining whether video content is popular with a community, features such as video theme, video quality, etc., have a direct impact on the results in addition to the title and cover features that are directly presented to the user. Therefore, the embodiment of the application provides a wide & deep classification network structure based on video multi-mode information, wherein a factor decomposition machine (FM, factorization Machine) is used for crossing low-order features such as video category, video duration, video quality and the like in the wide part, and multi-mode fusion is performed on high-order features such as a title, a label, a cover and the like in the deep part.

The network structure diagram of the overall algorithm in the embodiment of the application is shown in fig. 10, and the characteristics of network input are as follows: firstly, inputting low-order characteristics of a wide network, wherein the low-order characteristics comprise three types of categories to which videos belong, video duration and quality grades of the videos; and the other is to input three characteristics of a title, a label and a cover map of the deep network. The window and deep classification network model uses the 6 video features to obtain two paths of respective prediction scores through two paths of transformation of the window and deep respectively, and finally sums the two paths of prediction results, and transforms the prediction results into a [0,1] range by using a sigmoid nonlinear activation function to obtain the final popularity score.

In the model training phase of the wide & deep classification network model, the cross entropy loss function provided by the following equation (1-1) may be used to measure the cost of misclassification:

Wherein L represents the loss result of the cross entropy loss function; avg represents the average value; y _n is the prediction probability of the nth sample; y _n e {0,1} is the true trend result of the nth sample, where 0 represents unwelcome and 1 represents popular; in the model prediction stage, the prediction probability y output by the network is used, if y is more than 0.5, the video is judged to be popular, and otherwise, the video is judged to be unpopular.

Based on the network structure diagram provided in fig. 10, the left half part in fig. 10 represents a wide network structure, and the input features include three parts of category of video, video duration and video quality level. The main purpose of using the wide network is to introduce the memory capability of the linear model to amplify the impact of specific feature combinations in the trend identification task on the final decision. Such as { entertainment-the eight diagrams, high quality, moderate duration } video, the probability of being popular with the female population is high. Specifically, in the embodiment of the present application, an embedding layer 1101 is used to obtain an embedding vector of an input feature, an FM layer 1002 is used as a feature cross network, and different preprocessing methods are used for the three features of the input.

In the embodiment of the application, for the video category characteristics: the video can be classified into a certain category according to the video content, and the video is specifically classified into first-class classification and second-class classification, for example, 42 first-class classification and 296 second-class classification are combined, and single-heat coding (one hot) is respectively carried out on the first-class classification and the second-class classification, so that two category feature vectors x _cls1∈R⁴²,x_cls2∈R²⁹⁶ can be obtained;

For video duration feature: the video duration is a continuous characteristic, so that equal frequency bin discretization can be performed, the video duration is classified into 1-5 grades, and the video duration is expressed as x _t∈R⁵ by using one hot code;

For video quality features: the video quality can be scored according to prior standards such as video definition, video content and the like, and the scoring result is directly used in the embodiment of the application and classified into 1-5 grades, and the o ne hot code is also used as x _level∈R⁵.

Then, the four discrete features x _cls1、x_cls2、x_t and x _level are input into the FM layer, and the output y _wide of the wide portion can be obtained by the following formula (1-2).

y_wide＝FM(x_cls1,x_cls2,x_t,x_level) (1-2)。

With continued reference to the network structure diagram provided in fig. 10, the right half of fig. 10 shows a deep network structure, and the input features used in this part are divided into two modes, namely text and image, where the text part includes a title and a tag (tag) of a video, and the image part is a cover of the video. Since text and image information is cross-modal, embodiments of the present application also introduce inter-modal feature fusion based on an attention mechanism (attention) to better express video content features.

Firstly, on feature extraction, word segmentation is carried out on title information, word2vec is used for obtaining word embedded vectors (embedding) of each word, embedding results of all words are averaged, and then a title sentence level coding vector x _title is obtained; the tag information is a word sequence formed by key description words in the video, so that the word segmentation process is omitted, and after embedding coding results of each word are directly obtained, the average of all words embedding is calculated to obtain a coding vector x _tag of the tag; for the cover map, a resn et network (a depth residual error network, which is a CNN-based picture feature extraction network) is directly adopted for feature extraction in the embodiment of the application to obtain a cover map vector x _img.

After extracting each input feature, carrying out inter-mode feature fusion by using attention mechanism. The specific method comprises the steps of introducing alpha _title、α_tag、α_img attention weights into a network, respectively giving different weights to x _title、x_tag、x_img, respectively carrying out linear transformation on weighted results through three paths of Fully Connected networks (FC), then carrying out concat splicing to serve as a feature fusion result, and carrying out subsequent operation, wherein f _i (x) =Wx+b is a calculation process of concat splicing processing, wherein x represents an input feature vector of the FC layer, W and B are weight parameters to be learned by a model, and the FC layer represents linear transformation on the input feature vector.

X＝concat(f₁(α_title·x_title),f₂(α_tag·x_tag),f₃(α_img·x_img)) (1-3);

Wherein, X represents the result after concat splicing processing.

After the multi-mode fusion feature X is obtained, 3 layers of NN transformation are carried out, and finally the output dimension is reduced to 1 dimension, so that the popularity score y _deep =DNN (X) of the deep network prediction is obtained.

Finally, the output results of the wide network and the deep network are combined through the following formula (1-4), and the overall popularity score is obtained as y.

y＝sig moid(y_wide+y_deep) (1-4)。

In the embodiment of the application, the consumption log of the specific user group is used as the labeling data of whether the content is popular, the crowd tendency prediction problem is modeled as the two classification problems of whether the content is popular with the specific crowd, and the label is directly constructed by using the user operation data, so that a large amount of labeling cost is saved. By using a wide and deep network model structure based on multi-mode information, the complexity and diversity of characteristics under a tendency prediction scene are fully considered, so that the model can more accurately describe the popularity of contents on specific crowds. By utilizing crowd tendency prediction results of the content, the recommendation side can better solve the problem of cold start of a new user, and the retention and the duration of the user are improved; in addition, on the basis of optimizing the recommendation effect for a specific crowd, the result can provide a content pool popular with the crowd, and is important for improving the consumption index of the specific crowd.

Continuing with the description below of an exemplary structure in which the information recommendation device 354 provided by embodiments of the present application is implemented as a software module, in some embodiments, as shown in fig. 2, the software module stored in the information recommendation device 354 of the memory 350 may be an information recommendation device in the server 300, including:

An obtaining module 3541, configured to obtain at least one first type feature and at least one second type feature of each information to be recommended in the information to be recommended set; the feature cross processing module 3542 is configured to perform feature cross processing on the at least one first type of feature to obtain a first prediction score; the feature fusion processing module 3543 is configured to perform feature fusion processing on the at least one second class feature to obtain a second prediction score; the transformation processing module 3544 is configured to perform a prediction result transformation process on the first prediction score and the second prediction score, so as to obtain a tendency score of the information to be recommended; and a recommending module 3545, configured to recommend at least one piece of information to be recommended in the set of information to be recommended to a target object according to the tendency score of each piece of information to be recommended.

In some embodiments, the apparatus further comprises: a determining module, configured to determine attribute information of the target object; the collection acquisition module is used for acquiring an object collection formed by the objects with the attribute information and a recommendation information collection formed by recommendation information with interaction information between the objects in the object collection; the interactive data acquisition module is used for acquiring interactive data between each object in the object set and the recommended information for each recommended information in the recommended information set; the aggregation module is used for aggregating the interaction data to obtain interaction probability of the objects in the object set for each piece of recommended information; and the selection module is used for selecting at least one piece of recommended information from the recommended information set as information to be recommended according to the interaction probability so as to form the information set to be recommended.

In some embodiments, the interactive data includes at least exposure times, click times and browsing time periods for the recommended information; the aggregation module is further configured to: aggregating the exposure times, the clicking times and the browsing time length to obtain average clicking rate and average browsing time length for each piece of recommended information; and determining any one of the average click rate and the average browsing duration as the interaction probability of the recommended information.

In some embodiments, the apparatus further comprises: the total average click rate determining module is used for determining the total average click rate corresponding to all objects in the object set according to the average click rate of each piece of recommended information in the recommended information set; a total average browsing duration module, configured to determine a total average browsing duration corresponding to all objects in the object set according to the average browsing duration of each piece of recommended information in the recommended information set; the selection module is further configured to: selecting the recommended information with the average click rate larger than the total average click rate from the recommended information set as the information to be recommended; or selecting the recommended information with the average browsing time length longer than the total average browsing time length from the recommended information set as the information to be recommended.

In some embodiments, the first class of features includes at least one of an information class feature, an information length feature, and an information quality feature; the feature cross processing module is further configured to: respectively encoding the first class feature and the second class feature in the information class features to correspondingly obtain a first class feature vector and a second class feature vector; sequentially performing discretization and coding on the information length characteristics to obtain information length characteristic vectors; coding the information quality characteristics to obtain information quality characteristic vectors; and performing the cross processing on at least one of the primary category feature vector, the secondary category feature vector, the information length feature vector and the information quality feature vector by adopting a factor decomposition machine to obtain the first prediction score.

In some embodiments, the feature cross-processing module is further to: acquiring a pre-scoring value for the information to be recommended; and carrying out the coding processing on the pre-scoring value to obtain the information quality feature vector.

In some embodiments, the second type of features include at least text features and image features; the feature fusion processing module is further used for: determining at least one word from the text feature; acquiring word embedding vectors of each word; determining the average value of the word embedding vectors as the coding vector of the text feature; extracting the image features to obtain image feature vectors; and carrying out feature fusion processing on the coding vector and the image feature vector based on an attention mechanism to obtain the second prediction component.

In some embodiments, the feature fusion processing module is further configured to: based on the attention mechanism, respectively performing linear transformation on the coding vector and the image feature vector to correspondingly obtain a linearly transformed coding vector and a linearly transformed image feature vector; splicing the encoding vector after linear transformation and the image characteristic vector after linear transformation to obtain a spliced vector; and performing activation processing on the spliced vector to reduce the dimension of the spliced vector to one dimension, so as to obtain the second prediction score.

In some embodiments, the feature fusion processing module is further configured to: determining a first attention weight of the encoding vector and a second attention weight of the image feature vector; weighting the encoding vector and the image feature vector with the first attention weight and the second attention weight, respectively; and respectively carrying out linear transformation processing on the weighted coding vector and the weighted image characteristic vector to correspondingly obtain the linear transformed coding vector and the linear transformed image characteristic vector.

In some embodiments, the transformation processing module is further to: summing the first prediction score and the second prediction score to obtain a prediction total score; and carrying out prediction result transformation processing on the prediction total score by adopting a nonlinear activation function to obtain the tendency score of the information to be recommended.

In some embodiments, the apparatus further comprises: the processing module is used for determining a tendency score of the information to be recommended by adopting a tendency identification network; wherein the trend identification network is trained by: inputting a first type of sample features of sample information and a second type of sample features of the sample information into the trend identification network; performing feature cross processing on the first type sample features through a feature cross network of the tendency identification network to obtain a first sample prediction score; performing feature fusion processing on the second type sample features through a feature fusion network of the tendency identification network to obtain second sample prediction scores; performing prediction result transformation processing on the first sample prediction score and the second sample prediction score through a transformation processing layer of the tendency recognition network to obtain a sample tendency score of the sample information; inputting the sample tendency score into a preset loss model to obtain a loss result; and correcting parameters in the characteristic crossing network and the characteristic fusion network according to the loss result to obtain a trained tendency identification network.

It should be noted that, the description of the apparatus according to the embodiment of the present application is similar to the description of the embodiment of the method described above, and has similar beneficial effects as the embodiment of the method, so that a detailed description is omitted. For technical details not disclosed in the present apparatus embodiment, please refer to the description of the method embodiment of the present application for understanding.

Embodiments of the present application provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the method according to the embodiment of the present application.

Embodiments of the present application provide a storage medium having stored therein executable instructions which, when executed by a processor, cause the processor to perform a method provided by embodiments of the present application, for example, as shown in fig. 3.

In some embodiments, the storage medium may be a computer-readable storage medium, such as ferroelectric Memory (FRAM, ferromagnetic Random Access Memory), read-Only Memory (ROM, R ead Only Memory), programmable Read-Only Memory (PROM, programmable Read Only Memory), erasable programmable Read-Only Memory (EPROM, erasable Programmable Read Only Memory), electrically erasable programmable Read-Only Memory (EEPROM, ELECTRICALLY ERASABLE PROGRAMMABLE READ ONLY MEMORY), flash Memory, magnetic surface Memory, optical Disk, or Compact Disk-Read Only Memory (CD-ROM), among others; but may be a variety of devices including one or any combination of the above memories.

In some embodiments, the executable instructions may be in the form of programs, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and they may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

As an example, the executable instructions may, but need not, correspond to files in a file system, may be stored as part of a file that holds other programs or data, such as in one or more scripts in a hypertext markup language (html, hyper Text Markup Language) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub-programs, or portions of code). As an example, executable instructions may be deployed to be executed on one computing device or on multiple computing devices located at one site or distributed across multiple sites and interconnected by a communication network.

The foregoing is merely exemplary embodiments of the present application and is not intended to limit the scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. An information recommendation method, comprising:

Acquiring first type features and second type features of each piece of information to be recommended in an information set to be recommended, wherein the first type features comprise information category features, information length features and information quality features, and the second type features at least comprise text features and image features;

respectively encoding the first class feature and the second class feature in the information class features to correspondingly obtain a first class feature vector and a second class feature vector;

Sequentially performing discretization and coding on the information length characteristics to obtain information length characteristic vectors;

Acquiring a pre-scoring value for the information to be recommended;

Performing the coding processing on the pre-scoring value to obtain the information quality feature vector;

Performing cross processing on the primary category feature vector, the secondary category feature vector, the information length feature vector and the information quality feature vector by adopting a factor decomposition machine to obtain a first prediction component;

determining at least one word from the text feature;

acquiring word embedding vectors of each word;

determining the average value of the word embedding vectors as the coding vector of the text feature;

extracting the image features to obtain image feature vectors;

based on an attention mechanism, carrying out feature fusion processing on the coding vector and the image feature vector to obtain a second prediction score;

performing prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended;

And recommending at least one piece of information to be recommended in the information set to be recommended to a target object according to the tendency score of each piece of information to be recommended.

2. The method according to claim 1, wherein the method further comprises:

Determining attribute information of the target object;

Acquiring an object set formed by the objects with the attribute information and a recommendation information set formed by recommendation information with interaction information between the objects in the object set;

for each piece of recommended information in the recommended information set, acquiring interaction data between each object in the object set and the recommended information;

aggregating the interaction data to obtain interaction probability of the objects in the object set for each piece of recommended information;

And selecting at least one piece of recommended information from the recommended information set as information to be recommended according to the interaction probability so as to form the information set to be recommended.

3. The method of claim 2, wherein the interactive data includes at least a number of exposures, a number of clicks, and a browsing duration of the recommended information;

the aggregating the interaction data to obtain the interaction probability of the object in the object set for each recommended information includes:

Aggregating the exposure times, the clicking times and the browsing time length to obtain average clicking rate and average browsing time length for each piece of recommended information;

And determining any one of the average click rate and the average browsing duration as the interaction probability of the recommended information.

4. A method according to claim 3, characterized in that the method further comprises:

Determining the total average click rate corresponding to all objects in the object set according to the average click rate of each piece of recommended information in the recommended information set;

determining the total average browsing duration corresponding to all objects in the object set according to the average browsing duration of each piece of recommended information in the recommended information set;

According to the interaction probability, selecting at least one piece of recommended information from the recommended information set as information to be recommended, including:

selecting the recommended information with the average click rate larger than the total average click rate from the recommended information set as the information to be recommended; or alternatively

And selecting the recommended information with the average browsing time length longer than the total average browsing time length from the recommended information set as the information to be recommended.

5. The method according to claim 1, wherein the performing feature fusion processing on the encoded vector and the image feature vector based on the attention mechanism to obtain a second prediction score includes:

Based on the attention mechanism, respectively performing linear transformation on the coding vector and the image feature vector to correspondingly obtain a linearly transformed coding vector and a linearly transformed image feature vector;

Splicing the encoding vector after linear transformation and the image characteristic vector after linear transformation to obtain a spliced vector;

and performing activation processing on the spliced vector to reduce the dimension of the spliced vector to one dimension, so as to obtain the second prediction score.

6. The method according to claim 5, wherein the performing linear transformation on the encoded vector and the image feature vector based on the attention mechanism respectively, corresponds to obtaining a linearly transformed encoded vector and a linearly transformed image feature vector, includes:

determining a first attention weight of the encoding vector and a second attention weight of the image feature vector;

Weighting the encoding vector and the image feature vector with the first attention weight and the second attention weight, respectively;

And respectively carrying out linear transformation processing on the weighted coding vector and the weighted image characteristic vector to correspondingly obtain the linear transformed coding vector and the linear transformed image characteristic vector.

7. The method according to claim 1, wherein the performing a prediction result transformation process on the first prediction score and the second prediction score to obtain the tendency score of the information to be recommended includes:

Summing the first prediction score and the second prediction score to obtain a prediction total score;

And carrying out prediction result transformation processing on the prediction total score by adopting a nonlinear activation function to obtain the tendency score of the information to be recommended.

8. The method according to any one of claims 1 to 7, further comprising: determining a tendency score of the information to be recommended by adopting a tendency identification network;

Wherein the trend identification network is trained by:

inputting a first type of sample features of sample information and a second type of sample features of the sample information into the trend identification network;

Performing feature cross processing on the first type sample features through a feature cross network of the tendency identification network to obtain a first sample prediction score;

performing feature fusion processing on the second type sample features through a feature fusion network of the tendency identification network to obtain second sample prediction scores;

performing prediction result transformation processing on the first sample prediction score and the second sample prediction score through a transformation processing layer of the tendency recognition network to obtain a sample tendency score of the sample information;

Inputting the sample tendency score into a preset loss model to obtain a loss result;

And correcting parameters in the characteristic crossing network and the characteristic fusion network according to the loss result to obtain a trained tendency identification network.

9. An information recommendation device, characterized by comprising:

The information recommendation system comprises an acquisition module, a recommendation module and a recommendation module, wherein the acquisition module is used for acquiring first type features and second type features of each piece of information to be recommended in an information set, the first type features comprise information category features, information length features and information quality features, and the second type features at least comprise text features and image features;

The characteristic cross processing module is used for respectively carrying out coding processing on the first class characteristic and the second class characteristic in the information class characteristics to correspondingly obtain a first class characteristic vector and a second class characteristic vector; sequentially performing discretization and coding on the information length characteristics to obtain information length characteristic vectors; acquiring a pre-scoring value for the information to be recommended; performing the coding processing on the pre-scoring value to obtain the information quality feature vector; performing cross processing on the primary category feature vector, the secondary category feature vector, the information length feature vector and the information quality feature vector by adopting a factor decomposition machine to obtain a first prediction component;

The feature fusion processing module is used for determining at least one word according to the text features; acquiring word embedding vectors of each word; determining the average value of the word embedding vectors as the coding vector of the text feature; extracting the image features to obtain image feature vectors; based on an attention mechanism, carrying out feature fusion processing on the coding vector and the image feature vector to obtain a second prediction score;

The transformation processing module is used for carrying out prediction result transformation processing on the first prediction score and the second prediction score to obtain a tendency score of the information to be recommended;

And the recommending module is used for recommending at least one piece of information to be recommended in the information set to be recommended to a target object according to the tendency score of each piece of information to be recommended.

10. An information recommendation device, characterized by comprising:

a memory for storing executable instructions; a processor for implementing the information recommendation method according to any one of claims 1 to 8 when executing executable instructions stored in said memory.

11. A computer readable storage medium, characterized in that executable instructions are stored for causing a processor to execute the executable instructions for implementing the information recommendation method according to any one of claims 1 to 8.