Information recommendation method and device, electronic equipment and medium

Abstract

The invention provides an information recommendation method and device, electronic equipment, a computer readable storage medium and a computer program product, and relates to the field of artificial intelligence, in particular to the technical field of data processing, intelligent recommendation and deep learning. According to the implementation scheme, for each consumption time period in a plurality of consumption time periods, consumed information of a first user is acquired; determining a first feature vector of each piece of information, a second feature vector of the first user and a third feature vector of the to-be-recommended information; for each consumption time period, inputting the first feature vector in the consumption time period into a preset self-attention layer to obtain a fourth feature vector corresponding to the consumption time period; inputting the first feature vector combination into a preset neural network model to obtain a predicted value, wherein the first feature vector combination comprises a fourth feature vector, a second feature vector and a third feature vector; and determining whether to recommend the to-be-recommended information to the first user based on the predicted value.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and in particular to the fields of data processing, intelligent recommendation, and deep learning technology, and specifically to an information recommendation method, device, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] Artificial intelligence is a discipline that studies how to use computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing. Artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0003] With the development of the Internet, various online services and application software have become increasingly popular. While people enjoy the convenience brought by the growth of information, the explosive growth of information has caused serious information overload. Recommendation systems can analyze user historical behaviors such as browsing records and purchase records to infer user preferences and needs, and recommend appropriate items to users based on this. Summary of the invention

[0004] The present disclosure provides an information recommendation method, device, electronic device, computer-readable storage medium, and computer program product.

[0005] According to one aspect of the present disclosure, an information recommendation method is provided, comprising: for each consumption time period in a plurality of consumption time periods, obtaining a first information identifier of information consumed by a first user in the consumption time period; determining a first feature vector corresponding to each of the first information identifiers, and a second feature vector corresponding to the first user identifier corresponding to the first user; determining a third feature vector corresponding to the second information identifier of the information to be recommended; for each of the consumption time periods, inputting each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; inputting the first feature vector combination into a preset neural network model to obtain an output prediction value, wherein the prediction value is used to represent the first user's interest level in the information to be recommended, wherein the first feature vector combination includes: the fourth feature vector, the second feature vector and the third feature vector; and determining whether to recommend the information to be recommended to the first user based on the prediction value.

[0006] According to another aspect of the present disclosure, a model training method is provided, including: for each consumption time period in a plurality of consumption time periods, obtaining a first information identifier of information consumed by a first user in the consumption time period; determining a first feature vector corresponding to each of the first information identifiers, and a second feature vector corresponding to the first user identifier corresponding to the first user; determining a third feature vector corresponding to the second information identifier of information to be recommended, and preset label data for determining whether to recommend the information to be recommended to the first user; for each consumption time period, inputting each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; inputting a combination of the first feature vectors into a neural network model to obtain a predicted value, wherein the predicted value is used to represent the first user's interest in the information to be recommended, wherein the combination of the first feature vectors includes: a fourth feature vector, the second feature vector and the third feature vector; determining a loss value based on the predicted value and the label data by a preset loss function; and training the neural network model based on the loss value.

[0007] According to another aspect of the present disclosure, an information recommendation device is provided, comprising: a first acquisition unit, configured to acquire, for each consumption time period in a plurality of consumption time periods, a first information identifier of information consumed by a first user in the consumption time period; a first determination unit, configured to determine a first feature vector corresponding to each of the first information identifiers, and a second feature vector corresponding to a first user identifier corresponding to the first user; a second determination unit, configured to determine a third feature vector corresponding to the second information identifier of the information to be recommended; a first input unit, configured to input, for each of the consumption time periods, each of the first feature vectors acquired in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; a second input unit, configured to input the first feature vector combination into a preset neural network model to obtain an output prediction value, wherein the prediction value is used to represent the first user's interest level in the information to be recommended, wherein the first feature vector combination includes: the fourth feature vector, the second feature vector and the third feature vector; and a recommendation unit, configured to determine whether to recommend the information to be recommended to the first user based on the prediction value.

[0008] According to another aspect of the present disclosure, a model training device is provided, including: a second acquisition unit, configured to acquire, for each consumption time period in a plurality of consumption time periods, a first information identifier of information consumed by a first user in the consumption time period; a third determination unit, configured to determine a first feature vector corresponding to each of the information identifiers, and a second feature vector corresponding to the first user identifier corresponding to the first user; a fourth determination unit, configured to determine a third feature vector corresponding to the second information identifier of the information to be recommended, and preset label data for determining whether to recommend the information to be recommended to the first user; a third input unit, configured to input, for each consumption time period, each of the first feature vectors acquired in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; a fourth input unit, configured to input a combination of the first feature vectors into a neural network model to obtain a predicted value, wherein the predicted value is used to represent the first user's interest in the information to be recommended, wherein the combination of the first feature vectors includes: a fourth feature vector, the second feature vector and the third feature vector; a calculation unit, configured to determine a loss value through a preset loss function based on the predicted value and the label data; and an adjustment unit, configured to train the neural network model based on the loss value.

[0009] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in the present disclosure.

[0010] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided. The computer instructions are used to cause a computer to execute the method described in the present disclosure.

[0011] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the method described in the present disclosure when executed by a processor.

[0012] According to one or more embodiments of the present disclosure, the user's historical portrait is expanded by extracting user consumption information within multiple consumption time periods, and the feature extraction of the user's historical portrait in each time period is achieved through the self-attention layer, thereby solving the problem of users forgetting their long-term interests and improving the accuracy of information recommendation.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0015] Figure 1 A schematic diagram showing an exemplary system in which the various methods described herein may be implemented according to an embodiment of the present disclosure;

[0016] Figure 2 A flowchart of an information recommendation method according to an embodiment of the present disclosure is shown;

[0017] Figure 3 A schematic diagram showing information recommendation according to an embodiment of the present disclosure is shown;

[0018] Figure 4 A flowchart of model training according to an embodiment of the present disclosure is shown;

[0019] Figure 5 A structural block diagram of an information recommendation device according to an embodiment of the present disclosure is shown;

[0020] Figure 6 A structural block diagram of a model training device according to an embodiment of the present disclosure is shown; and

[0021] Figure 7 A structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0022] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0023] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0024] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0025] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0026] Figure 1 FIG. 1 is a schematic diagram of an exemplary system 100 in which various methods and apparatuses described herein may be implemented according to an embodiment of the present disclosure. Figure 1 , the system 100 includes one or more client devices 101, 102, 103, 104, 105, and 106, a server 120, and one or more communication networks 110 coupling the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105, and 106 may be configured to execute one or more applications.

[0027] In an embodiment of the present disclosure, the server 120 may run one or more services or software applications that enable the method of information recommendation or model training to be performed.

[0028] In some embodiments, server 120 may also provide other services or software applications, which may include non-virtualized environments and virtualized environments. In some embodiments, these services may be provided as web-based services or cloud services, such as provided to users of client devices 101, 102, 103, 104, 105, and / or 106 under a software as a service (SaaS) model.

[0029] exist Figure 1 In the configuration shown, the server 120 may include one or more components that implement the functions performed by the server 120. These components may include software components, hardware components, or a combination thereof that can be executed by one or more processors. Users operating client devices 101, 102, 103, 104, 105, and / or 106 may in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by these components. It should be understood that a variety of different system configurations are possible, which may differ from the system 100. Therefore, Figure 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

[0030] The user may use client devices 101, 102, 103, 104, 105 and / or 106 to obtain recommendation information. The client device may provide an interface that enables the user of the client device to interact with the client device. The client device may also output information to the user via the interface. Figure 1 Only six client devices are depicted, but one skilled in the art will appreciate that the present disclosure may support any number of client devices.

[0031] Client devices 101, 102, 103, 104, 105 and / or 106 may include various types of computer devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, game systems, thin clients, various messaging devices, sensors or other sensing devices, etc. These computer devices may run various types and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as GOOGLE Chrome OS); or include various mobile operating systems, such as MICROSOFT Windows Mobile OS, iOS, Windows Phone, Android. Portable handheld devices may include cellular phones, smart phones, tablet computers, personal digital assistants (PDAs), etc. Wearable devices may include head-mounted displays (such as smart glasses) and other devices. Game systems may include various handheld game devices, Internet-enabled game devices, etc. Client devices are capable of executing various different applications, such as various Internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and may use various communication protocols.

[0032] The network 110 may be any type of network known to those skilled in the art that may support data communications using any of a variety of available protocols, including but not limited to TCP / IP, SNA, IPX, etc. By way of example only, the one or more networks 110 may be a local area network (LAN), an Ethernet-based network, a token ring, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a blockchain network, a public switched telephone network (PSTN), an infrared network, a wireless network (e.g., Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0033] Server 120 may include one or more general purpose computers, dedicated server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and / or combination. Server 120 may include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that may be virtualized to maintain a server's virtual storage device). In various embodiments, server 120 may run one or more services or software applications that provide the functionality described below.

[0034] The computing units in the server 120 may run one or more operating systems including any of the above operating systems and any commercially available server operating systems. The server 120 may also run any of a variety of additional server applications and / or middle-tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0035] In some implementations, server 120 may include one or more applications to analyze and consolidate data feeds and / or event updates received from users of client devices 101, 102, 103, 104, 105, and 106. Server 120 may also include one or more applications to display data feeds and / or real-time events via one or more display devices of client devices 101, 102, 103, 104, 105, and 106.

[0036] In some embodiments, the server 120 may be a server of a distributed system, or a server combined with a blockchain. The server 120 may also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. A cloud server is a host product in a cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services.

[0037] The system 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as user information, information to be recommended, information consumed, etc. The database 130 may reside in various locations. For example, the database used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection. The database 130 may be of different types. In some embodiments, the database used by the server 120 may be, for example, a relational database. One or more of these databases may store, update, and retrieve data from the database in response to a command.

[0038] In some embodiments, one or more of the databases 130 may also be used by applications to store application data. The databases used by the applications may be different types of databases, such as a key-value store, an object store, or a conventional store backed by a file system.

[0039] Figure 1 The system 100 may be configured and operated in various ways to enable the application of various methods and apparatuses described in the present disclosure.

[0040] Usually, the ranking model in the recommendation system has a feature design based on user information and the content to be recommended, and the superimposed model training attenuates the training mode of long-term features, which makes the model more likely to remember short-term interests and gradually forget the user's long-term interests, causing the user's interests to fall into a cocoon state.

[0041] Therefore, according to an embodiment of the present disclosure, an information recommendation method is provided. Figure 2 A flowchart of an information recommendation method according to an embodiment of the present disclosure is shown. Figure 2As shown, method 200 includes: for each consumption time period in multiple consumption time periods, obtaining the first information identifier of the information consumed by the first user in the consumption time period (step 210); determining the first feature vector corresponding to each of the first information identifiers, and the second feature vector corresponding to the first user identifier corresponding to the first user (step 220); determining the third feature vector corresponding to the second information identifier of the information to be recommended (step 230); for each of the consumption time periods, inputting each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period (step 240); inputting the first feature vector combination into a preset neural network model to obtain an output prediction value, wherein the prediction value is used to represent the first user's interest in the information to be recommended, wherein the first feature vector combination includes: the fourth feature vector, the second feature vector and the third feature vector (step 250); and determining whether to recommend the information to be recommended to the first user based on the prediction value (step 260).

[0042] According to the embodiments of the present disclosure, the user's historical portrait is expanded by extracting user consumption information within multiple consumption time periods, and the feature extraction of the user's historical portrait in each time period is achieved through the self-attention layer, thereby solving the problem of users forgetting their long-term interests and improving the accuracy of information recommendation.

[0043] In the present disclosure, information may be text information, image information, audio information, video information, graphic information, product information, or any other suitable information for user consumption, without limitation.

[0044] In the present disclosure, "consumption" may refer to the action or operation of a user to obtain information through a communication medium. The system recommends information to the user through the communication medium. For example, the system pushes videos, pictures, texts and other information to the user through the user interface, which means that the video, picture, text and other information is recommended to the user; when the user browses the video, picture, text and other information through the user interface, the browsed information can be referred to as the consumed information. Alternatively, the system pushes product information to the user through the e-commerce platform, which means that the product information is recommended to the user; when the user purchases the product through the e-commerce platform, the purchased product information can be referred to as the consumed information.

[0045] In some examples, user historical consumption information may be obtained, and the information may have a content ID that uniquely represents the information. In some examples, the information identifier may include at least one of the content ID, the primary category of the content, the secondary category, the creator ID, and tag information, etc., without limitation.

[0046] In some examples, the first-level category, the second-level category and the tag information are used to identify the information respectively. For example, for a picture or text about food preparation, the first-level category may be food, the second-level category may be food preparation, and the tag information may be Sichuan cuisine, recipes, etc.

[0047] In the present disclosure, multiple consumption time periods can be determined according to consumption history time. For example, according to the user's consumption history time, the historical time is sequentially divided into short-term consumption time period, medium-term consumption time period, and long-term consumption time period from near to far. For example, the short-term consumption time period can be expressed as within the past 7 days, the medium-term consumption time period can be expressed as from 7 days ago to the previous 3 months, and the long-term consumption time period can be expressed as from 3 months ago to the previous 12 months.

[0048] In some embodiments, in step 210, for each consumption time period in a plurality of consumption time periods, a first information identifier of each of the one or more pieces of information consumed by the first user in the consumption time period is obtained.

[0049] According to some embodiments, obtaining the first information identifier of each of one or more information consumed by the first user during the consumption time period includes: for each information category in a plurality of information categories, obtaining the first information identifier of each of one or more information consumed by the first user during the consumption time period and under the information category.

[0050] In some examples, the information category can be a primary category or a secondary category. A secondary category refers to a more detailed category under a primary category. For example, in an art website, there may be a primary category "art", and then there may be several secondary categories under this level, such as "painting", "sculpture" and "photography". In this case, "painting" is a secondary category of the primary category "art".

[0051] According to some embodiments, the one or more information consumed is the one or more information that the first user is most interested in. The first user's interest in the consumed information is determined based on at least one of the following items: the cumulative number of clicks on the information by the first user in a corresponding consumption time period, and the single maximum consumption duration of the information by the first user in the consumption time period.

[0052] It is understandable that the information that the user is most interested in can better reflect the user's consumption profile information. In some examples, the first user's interest in the information can be characterized based on the cumulative number of clicks on the information by the first user in the corresponding consumption time period. For example, the more cumulative clicks a user has, the higher his or her interest level is.

[0053] In some examples, the first user's interest in the information can be characterized based on the first user's highest single consumption duration for the information within the consumption time period. Here, "single" can be a consumption process, such as a user browsing an article is a consumption process, and the browsing duration is the single consumption duration; when the user exits the article browsing interface and re-enters the article browsing interface, it is a new consumption process.

[0054] In some embodiments, for information such as video or audio, the maximum single consumption duration of the information by the first user during the consumption time period can be determined based on the user's browsing duration and the physical duration of the information such as video or audio. For example, the maximum single consumption duration can be normalized to the interval [0,1], such as the maximum single consumption duration can be determined by the ratio of the user's browsing duration to its physical duration. If the user's browsing duration is the same as its physical duration, the value is 1; if the user's browsing duration is 0, the value is 0.

[0055] In some examples, the first user's interest in the information may be characterized based on the cumulative number of clicks on the information by the first user in the corresponding consumption time period and the single maximum consumption duration of the first user on the information in the consumption time period. For example, the first user's interest in the information may be obtained by multiplying the cumulative number of clicks by the single maximum consumption duration.

[0056] It is understandable that the first user's interest level in the corresponding information may be determined in any suitable manner, which is not limited herein.

[0057] In some embodiments, in step 240, for each consumption time period, the first feature vector corresponding to each information obtained in the time period is input into a preset self-attention layer, so as to obtain multiple fourth feature vectors corresponding to the multiple consumption cycles one by one. According to some embodiments, the first user identifier includes a marking slot. Further, the method may also include: obtaining one or more second user identifiers by changing the marking slot of the first user identifier, wherein the number of user identifiers in the combination formed by the one or more second user identifiers and the first user identifier is the same as the number of the multiple consumption time periods; determining the fifth feature vector corresponding to each second user identifier of the one or more second user identifiers. At this time, the first feature vector combination also includes: the fifth feature vector corresponding to each second user identifier.

[0058] For example, assume that the first user identifier obtained is {240136699}:1, where the "1" after the ":" indicates its tag slot information. By changing the tag slot, one or more second user identifiers can be obtained, such as {240136699}:20001, {240136699}:30001. That is, the first user identifier and the second user identifier differ only in the tag slot information.

[0059] Through the above-mentioned operation of mirroring user identification, each consumption time period corresponds to a corresponding user identification, so that subsequent user characteristics can respectively characterize the user's portrait in each consumption time period, thereby strengthening the user's interest characteristics in each period.

[0060] According to some embodiments, the first feature vector combination further includes: a cross feature vector, wherein the cross feature vector is determined based on the first user identifier and tag information corresponding to the information to be recommended. The tag information is determined based on the content of the information to be recommended.

[0061] In the above embodiment, the tag information is determined based on the content of the information to be recommended. For example, for a graphic information introducing food preparation, the tag information may be: Sichuan cuisine, recipes, etc. The tag information may be multi-dimensional information, including multi-dimensional tags such as themes, topics, entities, etc., to cover key information in information such as articles.

[0062] In some examples, the tag information may be extracted through keyword extraction technology or based on deep learning technology, which is not limited here.

[0063] Through cross-feature design, the compatibility between the first user and the information to be recommended is further evaluated, thereby improving the accuracy of subsequent prediction values.

[0064] According to some embodiments, the preset neural network model includes: a densely connected network input layer, and a first fully connected layer. Inputting the first feature vector combination into the preset neural network model to obtain a predicted value includes: inputting the first feature vector combination into the densely connected network input layer to obtain a sixth feature vector; and obtaining the predicted value based on the sixth feature vector and through the first fully connected layer.

[0065] In some embodiments, obtaining the predicted value according to the sixth eigenvector and passing through the first fully connected layer includes: inputting the sixth eigenvector into the first fully connected layer to obtain the output predicted value. In some examples, when there are multiple first fully connected layers, the sixth eigenvector is sequentially input into the one or more first fully connected layers to obtain the predicted value.

[0066] In some examples, the first user's interest in the recommended information can be determined by the completion degree of the recommended information. The completion degree is an important indicator to measure the attractiveness of the recommended information content and the quality of user viewing. For example, the completion degree can be the predicted percentage of the first user who may watch the information content. For example, for a graphic information, the user may watch half of the graphic information, then the completion degree is 50%. For example, for a video information, the completion degree can be determined based on the ratio of the predicted user viewing time and the physical duration of the video. If the user viewing time = physical duration, the completion degree is 1.

[0067] In some examples, the densely connected network input layer and the first fully connected layer can form a densely connected network together. At this time, according to the core idea of ​​dense connection of the densely connected network, when multiple first fully connected layers are included, the input of the corresponding fully connected layer includes the output of all previous layers in the densely connected network in addition to the output of the previous layer. Through the densely connected network, the gradient disappearance is alleviated, the transmission of features is strengthened, so that features are more effectively utilized, and the number of parameters is reduced to a certain extent.

[0068] According to some embodiments, the preset neural network model also includes: a gate network, and a second fully connected layer. The method according to the present disclosure may also include: inputting the second feature vector combination into the gate network and the second fully connected layer in sequence to obtain the output seventh feature vector. At this time, according to the sixth feature vector and obtaining the predicted value through the first fully connected layer, it may also include: feature fusion of the sixth feature vector and the seventh feature vector to obtain an eighth feature vector; and inputting the eighth feature vector into the first fully connected layer to obtain the output predicted value. The second feature vector combination includes: other feature vectors in the first feature vector combination except the feature vector determined based on the information to be recommended.

[0069] In the above embodiment, the second feature vector combination includes other feature vectors in the first feature vector combination except the feature vector determined based on the information to be recommended. Specifically, when the first feature vector combination includes the plurality of fourth feature vectors, the second feature vector and the third feature vector, the second feature vector combination includes the plurality of fourth feature vectors and the second feature vector. When the first feature vector combination includes the plurality of fourth feature vectors, the second feature vector, the third feature vector and the fifth feature vector corresponding to each second user identifier (or further includes a cross feature vector), the second feature vector combination includes the plurality of fourth feature vectors, the second feature vector and the fifth feature vector corresponding to each second user identifier.

[0070] In some embodiments, the gate network may include one or more fully connected layers. Alternatively, the gate network includes one or more fully connected layers, sigmoid function layers and / or relu function layers.

[0071] Through the above embodiments, the user characteristics and the consumption portrait characteristics of the user in each consumption time period are strengthened, and a personalized bias structure is added to the consumption portrait characteristics of the user in different consumption time periods. The output characteristics are then connected to the main network (that is, fused with the output characteristics of the densely connected network input layer), so as to achieve the effect of strengthening the user's interest in each consumption time period, and then achieve accurate recommendations for users.

[0072] Figure 3 A schematic diagram of information recommendation according to an embodiment of the present disclosure is shown, Figure 3 As shown, s1, s2…s5k represent the first feature vectors of the information consumed by the first user in each consumption time period. The first feature vectors of the user consumption information in the long-term, short-term and medium-term periods are input into the serialized feature extraction module (i.e., the self-attention layer) to obtain three fourth feature vectors. The second feature vector (uid-1-0) corresponding to the first user and the two fifth feature vectors (uid-1-1, uid-1-2) obtained by mirroring, the third feature vector (nid) of the information to be recommended this time, the cross feature vector and the three fourth feature vectors are input into the input layer of the Dense network (dense connection network) to obtain the sixth feature vector. The three fourth feature vectors and the second feature vector (uid-1-0) corresponding to the first user and the two fifth feature vectors (uid-1-1, uid-1-2) obtained by mirroring are sequentially input into the gate network and the fully connected layer of the personalized bias structure to obtain the seventh feature vector; after the sixth feature vector and the seventh feature vector are subjected to feature fusion (such as inner product operation), the output eighth feature vector is sequentially input into the fully connected layer of the Dense network to obtain the output prediction value. Thereby, it is achieved to determine whether to recommend the information to be recommended to the first user based on the predicted value.

[0073] According to the embodiments of the present disclosure, Figure 4As shown, a model training method 400 is also provided, including: for each consumption time period in a plurality of consumption time periods, obtaining a first information identifier of information consumed by a first user in the consumption time period (step 410); determining a first feature vector corresponding to each of the first information identifiers, and a second feature vector corresponding to the first user identifier corresponding to the first user (step 420); determining a third feature vector corresponding to the second information identifier of the information to be recommended, and preset label data for determining whether to recommend the information to be recommended to the first user (step 430); for each consumption time period, inputting each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period (step 440); inputting the first feature vector combination into a neural network model to obtain a predicted value, wherein the predicted value is used to represent the first user's interest in the information to be recommended, wherein the first feature vector combination includes: a fourth feature vector, the second feature vector and the third feature vector (step 450); determining a loss value based on the predicted value and the label data through a preset loss function (step 460); and training the neural network model based on the loss value (step 470).

[0074] In some embodiments, the first user's interest in the recommended information can be determined by the completion degree of the information to be recommended. The completion degree is an important indicator to measure the attractiveness of the content of the information to be recommended and the quality of user viewing. For example, the completion degree can be the predicted percentage of the first user who may watch the content of the information. For example, for a graphic information, the user may watch half of the graphic information, then the completion degree is 50%. For example, for a video information, the completion degree can be determined based on the ratio of the predicted user viewing time and the physical duration of the video. If the user viewing time = physical duration, the completion degree is 1. Therefore, in some embodiments, the tag data can be a preset completion degree of the information to be recommended.

[0075] In some embodiments, in step 410, for each consumption time period in a plurality of consumption time periods, a first information identifier of each of the one or more information consumed by the first user in the consumption time period is obtained.

[0076] According to some embodiments, obtaining the first information identifier of each of one or more information consumed by the first user during the consumption time period includes: for each information category in a plurality of information categories, obtaining the first information identifier of each of one or more information consumed by the first user during the consumption time period and under the information category.

[0077] According to some embodiments, the one or more information consumed is the one or more information in which the first user is most interested, wherein the first user's interest in the consumed information is determined based on at least one of the following items: the cumulative number of clicks on the information by the first user within the corresponding consumption time period, and the single highest consumption duration of the information by the first user within the consumption time period.

[0078] In some embodiments, in step 440, for each consumption time period, the first feature vector corresponding to each information obtained within the time period is input into a preset self-attention layer, thereby obtaining multiple fourth feature vectors corresponding one-to-one to the multiple consumption cycles.

[0079] According to some embodiments, the first user identifier includes a marking slot, and wherein the method further includes: obtaining one or more second user identifiers by changing the marking slot of the first user identifier, wherein the number of user identifiers in a combination formed by the one or more second user identifiers and the first user identifier is the same as the number of the multiple consumption time periods; determining a fifth feature vector corresponding to each of the one or more second user identifiers; wherein the first feature vector combination further includes: a fifth feature vector corresponding to each of the second user identifiers.

[0080] According to some embodiments, the first feature vector combination further includes: a cross feature vector, wherein the cross feature vector is determined based on the first user identifier and tag information corresponding to the information to be recommended, wherein the tag information is determined based on the content of the information to be recommended.

[0081] According to some embodiments, the neural network model includes: a densely connected network input layer, and a first fully connected layer, wherein inputting the first feature vector combination into the neural network model to obtain a prediction value includes: inputting the first feature vector combination into the densely connected network input layer to obtain a sixth feature vector; and obtaining the prediction value based on the sixth feature vector and through the first fully connected layer.

[0082] In some embodiments, obtaining the predicted value according to the sixth eigenvector and passing through the first fully connected layer includes: inputting the sixth eigenvector into the first fully connected layer to obtain the output predicted value. In some examples, when there are multiple first fully connected layers, the sixth eigenvector is sequentially input into the one or more first fully connected layers to obtain the predicted value.

[0083] According to some embodiments, the neural network model further includes: a gate network, and a second fully connected layer, wherein the method further includes: inputting the second feature vector combination into the gate network and the second fully connected layer in sequence to obtain an output seventh feature vector. Obtaining the predicted value according to the sixth feature vector and through the first fully connected layer may also include: performing feature fusion on the sixth feature vector and the seventh feature vector to obtain an eighth feature vector; and inputting the eighth feature vector into the first fully connected layer to obtain the predicted value. The second feature vector combination includes: other feature vectors in the first feature vector combination except the feature vector determined based on the information to be recommended.

[0084] According to some embodiments, training the neural network model based on the loss value includes: based on the loss value, adjusting the parameters of the neural network model respectively through multiple optimizers, wherein the number of the multiple optimizers is the same as the number of the multiple consumption time periods.

[0085] According to some embodiments, the multiple consumption time periods include: short-term consumption time period, medium-term consumption time period, and long-term consumption time period determined according to consumption history time. The multiple optimizers include: an optimizer with a decay factor and an optimizer with automatic decay learning rate.

[0086] Therefore, according to some embodiments, the plurality of optimizers include: an RMSProp optimizer, an AdaGrad optimizer, and an Adam optimizer.

[0087] It is understandable that the optimizer with the attenuation factor converges faster, which is conducive to strengthening the importance of recent samples; the optimizer with the automatic attenuation learning rate design has a higher initial learning rate and converges faster, and the medium-term becomes smoother, so there is little difference in the use of medium-term and recent samples, which can help remember medium-term interests. In addition, in order to warm-start the model and use longer-term samples, the Adam optimizer is used. The Adam optimizer is currently a more stable and general optimizer.

[0088] It can be understood that here, in the process of backpropagation parameter update, the attenuation factors (or learning rates) of different optimizers are designed differently, and the updated parameter values ​​corresponding to the respective feature vectors are obtained through different optimizers, so that the user features controlled by multiple optimizers can express user interests of different periods.

[0089] According to some embodiments, determining the loss value based on the predicted value and the label data through a preset loss function includes: determining a first loss value based on the predicted value and the label data through a preset first loss function; determining a second loss value based on the second eigenvector and the fifth eigenvector through a preset second loss function. Adjusting the parameters of the neural network model based on the loss value includes: training the neural network model based on the first loss value and the second loss value so as to minimize the first loss value and make the second loss value greater than a preset threshold.

[0090] In some examples, when there are multiple second user identifiers, a second loss value is determined by a preset second loss function based on the second feature vector and fifth feature vectors corresponding to the multiple second user identifiers.

[0091] According to the above, the second feature vector and the fifth feature vector corresponding to each second user identifier are both determined based on the first user identifier. In order to achieve further differentiation, vector distance constraints are added between the feature vectors so that the distance between the feature vectors is as large as possible, thereby realizing the separate representation of multiple consumption time periods of user interests.

[0092] According to some embodiments, the preset second loss function is determined based on the following formula:

[0093] ||HH T -I|| F →Ω

[0094] Wherein, H represents the vector matrix composed of the second eigenvector and the fifth eigenvector, I represents the identity matrix, ||·|| F represents the matrix norm, and Ω represents the preset threshold.

[0095] In some examples, when there are multiple second user identifiers, H represents a vector matrix composed of the second eigenvector and the fifth eigenvectors corresponding to the multiple second user identifiers. That is, if there are two second user identifiers, the second eigenvector and two fifth eigenvectors (corresponding to the corresponding second user identifiers) constitute the vector matrix H.

[0096] It is understandable that the form of the above loss function is merely exemplary and is not intended to be limiting.

[0097] In the present disclosure, the model trained by the model training method described in any one of the above embodiments can be used to implement the information recommendation method described in any one of the embodiments of the present disclosure.

[0098] Here, the embodiments for implementing the model training method and the embodiments for implementing the information recommendation method, wherein the corresponding operations and features are similar, are not described in detail here.

[0099] According to the embodiments of the present disclosure, Figure 5 As shown, an information recommendation device 500 is also provided, including: a first acquisition unit 510, configured to acquire, for each consumption time period in a plurality of consumption time periods, a first information identifier of information consumed by a first user in the consumption time period; a first determination unit 520, configured to determine a first feature vector corresponding to each of the first information identifiers, and a second feature vector corresponding to a first user identifier corresponding to the first user; a second determination unit 530, configured to determine a third feature vector corresponding to the second information identifier of the information to be recommended; a first input unit 540, configured to, for each of the consumption time periods, input each of the first feature vectors acquired in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; a second input unit 550, configured to input the first feature vector combination into a preset neural network model to obtain an output prediction value, wherein the prediction value is used to represent the first user's interest in the information to be recommended, wherein the first feature vector combination includes: the fourth feature vector, the second feature vector and the third feature vector; and a recommendation unit 560, configured to determine whether to recommend the information to be recommended to the first user based on the prediction value.

[0100] Here, the operations of the above-mentioned units 510 to 560 of the information recommendation device 500 are similar to the operations of the above-described steps 210 to 260, and will not be described in detail here.

[0101] According to the embodiments of the present disclosure, Figure 6As shown, a model training device 600 is also provided, including: a second acquisition unit 610, configured to obtain, for each consumption time period in a plurality of consumption time periods, a first information identifier of information consumed by a first user in the consumption time period; a third determination unit 620, configured to determine a first feature vector corresponding to each of the information identifiers, and a second feature vector corresponding to a first user identifier corresponding to the first user; a fourth determination unit 630, configured to determine a third feature vector corresponding to a second information identifier of the information to be recommended, and preset label data for determining whether to recommend the information to be recommended to the first user; a third input unit 640, configured to, for each of the consumption time periods, Each of the first feature vectors obtained within the consumption time period is input into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; a fourth input unit 650 is configured to input the first feature vector combination into a neural network model to obtain a prediction value, wherein the prediction value is used to represent the first user's interest in the information to be recommended, wherein the first feature vector combination includes: a fourth feature vector, the second feature vector and the third feature vector; a calculation unit 660 is configured to determine a loss value through a preset loss function based on the prediction value and the label data; and an adjustment unit 670 is configured to train the neural network model based on the loss value.

[0102] Here, the operations of the above-mentioned units 610-670 of the model training device 600 are similar to the operations of the steps 410-470 described above, and will not be repeated here.

[0103] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0104] According to an embodiment of the present disclosure, an electronic device, a readable storage medium and a computer program product are also provided.

[0105] refer to Figure 7, a block diagram of an electronic device 700 that can be used as a server or client of the present disclosure will now be described, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0106] like Figure 7 As shown, the electronic device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the electronic device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0107] Multiple components in the electronic device 700 are connected to the I / O interface 705, including: an input unit 706, an output unit 707, a storage unit 708, and a communication unit 709. The input unit 706 can be any type of device that can input information to the electronic device 700. The input unit 706 can receive input digital or character information and generate key signal input related to user settings and / or function control of the electronic device, and can include but is not limited to a mouse, a keyboard, a touch screen, a track pad, a track ball, a joystick, a microphone, and / or a remote controller. The output unit 707 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 708 can include but is not limited to a disk, an optical disk. The communication unit 709 allows the electronic device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, an 802.11 device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0108] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as method 200 or 400. For example, in some embodiments, method 200 or 400 may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the method 200 or 400 described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the method 200 or 400 in any other appropriate manner (eg, by means of firmware).

[0109] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0110] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0111] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0112] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0113] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0114] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0115] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0116] Although the embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above-mentioned methods, systems and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but only by the claims after authorization and their equivalent scope. Various elements in the embodiments or examples can be omitted or replaced by their equivalent elements. In addition, each step can be performed in an order different from that described in the present disclosure. Further, the various elements in the embodiments or examples can be combined in various ways. It is important that with the evolution of technology, many elements described herein can be replaced by equivalent elements that appear after the present disclosure.

Claims

1. An information recommendation method, comprising: For each consumption time period in the plurality of consumption time periods, obtaining a first information identifier of information consumed by the first user in the consumption time period; Determine a first feature vector corresponding to each of the first information identifiers and a second feature vector corresponding to a first user identifier corresponding to the first user; Determine a third feature vector corresponding to the second information identifier of the information to be recommended; For each consumption time period, input each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; Inputting the first feature vector combination into a preset neural network model to obtain an output prediction value, wherein the prediction value is used to represent the first user's interest level in the information to be recommended, wherein the first feature vector combination includes: the fourth feature vector, the second feature vector and the third feature vector; and Determine whether to recommend the information to be recommended to the first user based on the predicted value.

2. The method of claim 1, wherein: The step of obtaining the first information identifier of the information consumed by the first user within the consumption time period includes: For each information category of the multiple information categories, the first information identifier of the first user in the consumption time period and under the information category is obtained.

3. The method according to claim 1 or 2, wherein: The consumed information is the information in which the first user has the highest interest, wherein the interest level is determined based on at least one of the following items: The cumulative number of clicks on the information by the first user in the corresponding consumption time period, and the maximum single consumption duration of the information by the first user in the consumption time period.

4. The method of claim 1, wherein: The first user identifier includes a marking slot, and wherein the method further includes: by changing the marking slot of the first user identifier to obtain a second user identifier, wherein the number of user identifiers in a combination of the second user identifier and the first user identifier is the same as the number of the plurality of consumption time periods; Determine a fifth feature vector corresponding to each of the second user identifiers; Wherein, the first eigenvector combination also includes: the fifth eigenvector.

5. The method according to claim 1 or 4, wherein: The first feature vector combination further includes: a cross feature vector, wherein the cross feature vector is determined based on the first user identifier and label information corresponding to the information to be recommended, wherein the label information is determined based on the content of the information to be recommended.

6. The method according to claim 1 or 4, wherein: The preset neural network model includes: a densely connected network input layer, and a first fully connected layer; The step of inputting the first feature vector combination into a preset neural network model to obtain an output prediction value comprises: Inputting the first feature vector combination into the densely connected network input layer to obtain an output sixth feature vector; and The predicted value is obtained according to the sixth eigenvector and through the first fully connected layer.

7. The method of claim 6, wherein: The preset neural network model also includes: a gate network, and a second fully connected layer; The method further comprises: The second feature vector combination is sequentially input into the gate network and the second fully connected layer to obtain the output seventh feature vector, and wherein, The obtaining the predicted value according to the sixth eigenvector and through the first fully connected layer includes any one of the following: Inputting the sixth feature vector into the first fully connected layer to obtain the output predicted value; The sixth eigenvector and the seventh eigenvector are subjected to feature fusion to obtain an eighth eigenvector; the eighth eigenvector is input into the first fully connected layer to obtain the output predicted value, The second feature vector combination includes: other feature vectors in the first feature vector combination except the feature vector determined based on the information to be recommended.

8. A model training method, comprising: For each consumption time period in the plurality of consumption time periods, obtaining a first information identifier of information consumed by the first user in the consumption time period; Determine a first feature vector corresponding to each of the first information identifiers and a second feature vector corresponding to a first user identifier corresponding to the first user; Determine a third feature vector corresponding to a second information identifier of the information to be recommended, and preset label data for determining whether to recommend the information to be recommended to the first user; For each consumption time period, input each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; Inputting the first feature vector combination into the neural network model to obtain a prediction value, wherein the prediction value is used to represent the first user's interest level in the information to be recommended, wherein the first feature vector combination includes: a fourth feature vector, the second feature vector and the third feature vector; Based on the predicted value and the label data, determining a loss value by using a preset loss function; and The neural network model is trained based on the loss value.

9. The method of claim 8, wherein: The obtaining of the first information identifier of the information consumed by the first user in the consumption time period includes: for each information category in a plurality of information categories, obtaining the first information identifier of the first user in the consumption time period and under the information category.

10. The method according to claim 8 or 9, wherein: The consumed information is the information in which the first user has the highest interest, wherein the interest level is determined based on at least one of the following items: The cumulative number of clicks on the information by the first user in the corresponding consumption time period, and the maximum single consumption duration of the information by the first user in the consumption time period.

11. The method of claim 8, wherein: The first user identifier includes a marking slot, and wherein the method further includes: by changing the marking slot of the first user identifier to obtain a second user identifier, wherein the number of user identifiers in a combination of the second user identifier and the first user identifier is the same as the number of the consumption time periods; Determine a fifth feature vector corresponding to each of the second user identifiers; Wherein, the first eigenvector combination also includes: the fifth eigenvector.

12. The method according to claim 8 or 11, wherein: The first feature vector combination further includes: a cross feature vector, wherein the cross feature vector is determined based on the first user identifier and label information corresponding to the information to be recommended, wherein the label information is determined based on the content of the information to be recommended.

13. The method according to claim 8 or 11, wherein: The neural network model includes: a densely connected network input layer, and a first fully connected layer; The step of inputting the first feature vector combination into the neural network model to obtain the output prediction value comprises: Inputting the first feature vector combination into the densely connected network input layer to obtain an output sixth feature vector; and The predicted value is obtained according to the sixth eigenvector and through the first fully connected layer.

14. The method of claim 13, wherein: The neural network model also includes: a gate network and a second fully connected layer; The method further comprises: The second feature vector combination is sequentially input into the gate network and the second fully connected layer to obtain the output seventh feature vector, and wherein, The obtaining the predicted value according to the sixth eigenvector in sequence and through the first fully connected layer includes any one of the following: Inputting the sixth feature vector into the first fully connected layer to obtain the output predicted value; The sixth eigenvector and the seventh eigenvector are subjected to feature fusion to obtain an eighth eigenvector; the eighth eigenvector is input into the first fully connected layer to obtain the output predicted value, The second feature vector combination includes: other feature vectors in the first feature vector combination except the feature vector determined based on the information to be recommended.

15. The method according to any one of claims 8 to 14, wherein: Adjusting the parameters of the neural network model based on the loss value includes: Based on the loss value, the neural network model is trained respectively by multiple optimizers, wherein the number of the multiple optimizers is the same as the number of the multiple consumption time periods.

16. The method of claim 15, wherein: The multiple consumption time periods include: short-term consumption time period, medium-term consumption time period, and long-term consumption time period determined according to consumption history time, and wherein the multiple optimizers include: an optimizer with a decay factor and an optimizer with automatic decay learning rate.

17. The method of claim 16, wherein: The multiple optimizers include: RMSProp optimizer, AdaGrad optimizer and Adam optimizer.

18. The method of claim 11, wherein: The determining the loss value based on the predicted value and the label data by using a preset loss function comprises: Based on the predicted value and the label data, determining a first loss value by using a preset first loss function; Based on the second eigenvector and the fifth eigenvector, a second loss value is determined by a preset second loss function, The training of the neural network model based on the loss value comprises: The neural network model is trained based on the first loss value and the second loss value so as to minimize the first loss value and make the second loss value greater than a preset threshold.

19. The method of claim 18, wherein: The preset second loss function is determined based on the following formula: ||HH T -I|| F →Ω Wherein, H represents a vector matrix composed of the second eigenvector and the fifth eigenvector, I represents an identity matrix, ||·|| F represents the matrix norm, and Ω represents the preset threshold.

20. An information recommendation device, comprising: A first acquisition unit is configured to acquire, for each consumption time period in the plurality of consumption time periods, a first information identifier of information consumed by the first user in the consumption time period; A first determining unit is configured to determine a first feature vector corresponding to each of the first information identifiers and a second feature vector corresponding to a first user identifier corresponding to the first user; A second determining unit, configured to determine a third feature vector corresponding to a second information identifier of the information to be recommended; A first input unit is configured to input each of the first feature vectors obtained in the consumption time period into a preset self-attention layer to obtain an output fourth feature vector corresponding to the consumption time period; a second input unit configured to input the first feature vector combination into a preset neural network model to obtain an output prediction value, wherein the prediction value is used to represent the first user's interest level in the information to be recommended, wherein the first feature vector combination includes: the fourth feature vector, the second feature vector and the third feature vector; and A recommendation unit is configured to determine whether to recommend the information to be recommended to the first user based on the predicted value.

21. A model training device, comprising: A second acquisition unit is configured to acquire, for each consumption time period in the plurality of consumption time periods, a first information identifier of information consumed by the first user in the consumption time period; A third determining unit is configured to determine a first feature vector corresponding to each of the information identifiers and a second feature vector corresponding to a first user identifier corresponding to the first user; a fourth determining unit configured to determine a third feature vector corresponding to the second information identifier of the information to be recommended, and preset label data for determining whether to recommend the information to be recommended to the first user; A third input unit is configured to input each of the first feature vectors obtained in the consumption time period into a preset self-attention layer for each consumption time period, so as to obtain an output fourth feature vector corresponding to the consumption time period; a fourth input unit, configured to input the first feature vector combination into the neural network model to obtain a prediction value, wherein the prediction value is used to represent the first user's interest level in the information to be recommended, wherein the first feature vector combination includes: a fourth feature vector, the second feature vector and the third feature vector; a calculation unit configured to determine a loss value by a preset loss function based on the predicted value and the label data; and An adjustment unit is configured to train the neural network model based on the loss value.

22. An electronic device, comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 19.

23. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-19.

24. A computer program product comprising a computer program, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 19 is implemented.

Images