CN117216393A

CN117216393A - Information recommendation method, training method and device of information recommendation model and equipment

Info

Publication number: CN117216393A
Application number: CN202311206948.8A
Authority: CN
Inventors: 袁鹏; 徐松; 赵靖; 黄蝶; 吴友政
Original assignee: Jingdong Technology Information Technology Co Ltd
Current assignee: Jingdong Technology Information Technology Co Ltd
Priority date: 2023-09-19
Filing date: 2023-09-19
Publication date: 2023-12-12

Abstract

The disclosure provides an information recommendation method, a training method of an information recommendation model, a training device of the information recommendation model and equipment, and the training method and the training device can be applied to the technical fields of artificial intelligence, deep learning, natural language processing and information recommendation. The information recommendation method comprises the following steps: acquiring user information of a user in response to detecting an operation behavior of the user aiming at a target link, wherein the user information comprises a historical browsing track information set and a user profile information set; processing the historical browsing track information set to obtain a historical browsing track feature sequence; processing the user profile information set to obtain a user profile feature sequence; and determining recommendation information corresponding to the user according to the historical browsing track feature sequence and the user profile feature sequence.

Description

Information recommendation method, training method and device of information recommendation model and equipment

Technical Field

The present disclosure relates to the technical fields of artificial intelligence, deep learning, natural language processing and information recommendation, and more particularly, to an information recommendation method, a training method and apparatus of an information recommendation model, an electronic device, a computer-readable storage medium and a computer program product.

Background

With the development of computer technology, intelligent customer service is increasingly widely used. Intelligent customer service may refer to an industry-application oriented technical means developed on the basis of large-scale knowledge processing.

The intelligent customer service integrates a large-scale knowledge processing technology, a natural language understanding technology, a knowledge management technology, an automatic question-answering system, an reasoning technology and the like. The problem recommendation of the intelligent customer service can refer to that before a user enters the customer service and asks the customer service, a recommendation system provides a problem list which the user possibly asks according to user information, so that the user can ask questions conveniently and the problems are more standard.

In the process of implementing the disclosed concept, the inventor finds that at least the following problems exist in the related art: since the recommendation method in the related art lacks learning ability for text information, accuracy and interpretability of the recommendation information cannot be ensured.

Disclosure of Invention

In view of this, the present disclosure provides an information recommendation method, an information recommendation model training method and apparatus, an electronic device, a computer readable storage medium, and a computer program product.

According to one aspect of the present disclosure, there is provided an information recommendation method including:

Acquiring user information of a user in response to detecting an operation behavior of the user aiming at a target link, wherein the user information comprises a historical browsing track information set and a user profile information set;

processing the historical browsing track information set to obtain a historical browsing track feature sequence;

processing the user profile information set to obtain a user profile feature sequence; and

and determining recommendation information corresponding to the user according to the historical browsing track feature sequence and the user profile feature sequence.

According to an embodiment of the present disclosure, the determining recommendation information corresponding to the user according to the historical browsing trajectory feature sequence and the user profile feature sequence includes:

processing the historical browsing track feature sequence and the user profile feature sequence to obtain an intermediate feature vector set; and

and processing the intermediate feature vector set to obtain the recommendation information.

According to an embodiment of the present disclosure, the historical browsing track feature sequence includes M historical browsing track feature vectors, the user profile feature sequence includes N user profile feature vectors, and M and N are positive integers.

According to an embodiment of the present disclosure, the processing the historical browsing trajectory feature sequence and the user profile feature sequence to obtain an intermediate feature vector set includes:

and performing feature cross processing on the M historical browsing track feature vectors and the N user profile feature vectors to obtain P explicit feature vectors and Q implicit feature vectors, wherein P and Q are positive integers.

According to an embodiment of the present disclosure, the processing the intermediate feature vector set to obtain the recommendation information includes:

for each explicit feature vector in the P explicit feature vectors, respectively performing fusion processing on the explicit feature vector and the Q implicit feature vectors to obtain Q target feature vectors; and

and determining the recommendation information according to Q target feature vectors corresponding to the P explicit feature vectors.

According to an embodiment of the present disclosure, performing feature cross processing on the M historical browsing trajectory feature vectors and the N user profile feature vectors to obtain P explicit feature vectors and Q implicit feature vectors includes:

for each historical browsing track feature vector in the M historical browsing track feature vectors, for each user profile feature vector in the N user profile feature vectors, performing low-order feature cross processing on the historical browsing track feature vector and the user profile feature vector to obtain the explicit feature vector; and

And performing high-order feature cross processing on the historical browsing track feature vector and the user profile feature vector to obtain the implicit feature vector.

According to an embodiment of the present disclosure, the above-described user profile information set includes a plurality of user profile information.

According to an embodiment of the present disclosure, the processing the user profile information set to obtain a user profile feature sequence includes:

determining a profile information type of the user profile information based on the user profile information, with respect to the user profile information among the plurality of user profile information;

when the profile information type is a continuous value type, determining a predetermined continuous coding mode corresponding to the continuous value type; and

and encoding the user profile information based on the predetermined continuous encoding mode to obtain a first user profile feature corresponding to the user profile information.

According to an embodiment of the present disclosure, the above information recommendation further includes:

when the profile information type is a discrete category type, determining a predetermined discrete coding mode corresponding to the discrete category type; and

and performing encoding processing on the user profile information based on the predetermined discrete encoding mode to obtain a second user profile characteristic corresponding to the user profile information.

According to an embodiment of the present disclosure, the set of historical browsing trajectory information includes at least one historical browsing trajectory information and association relationship information between the at least one historical browsing trajectory information.

According to an embodiment of the present disclosure, the processing the historical browsing track information set to obtain a historical browsing track feature sequence includes:

according to the association relation information, encoding the at least one historical browsing track information to obtain at least one position code and a historical browsing track feature vector corresponding to the at least one position code; and

and determining the historical browsing track feature sequence according to the at least one position code and the historical browsing track feature vector corresponding to the at least one position code.

According to an embodiment of the present disclosure, the obtaining, in response to detecting an operation behavior of a user with respect to a target link, user information corresponding to the user includes:

determining a user identification corresponding to the user in response to detecting the operation behavior of the user for the target link; and

and acquiring the historical browsing track information set and the user profile information set from a data source according to the user identification.

According to an embodiment of the present disclosure, the method further includes, after determining the recommendation information corresponding to the user according to the historical browsing trajectory feature sequence and the user profile feature sequence:

and displaying the recommendation information in an output area of the target page.

According to one aspect of the present disclosure, there is provided a training method of an information recommendation model, including:

acquiring sample user information corresponding to a sample user, wherein the sample user information comprises a sample historical browsing track information set, a sample user profile information set and real recommendation information;

processing the sample history browsing track information set to obtain a sample history browsing track characteristic sequence;

processing the sample user profile information set to obtain a sample user profile feature sequence;

determining first sample recommendation information corresponding to the sample user according to the sample historical browsing track feature sequence and the sample user profile feature sequence; and

and training a deep learning model by using the first sample recommendation information and the real recommendation information to obtain an information recommendation model.

According to an embodiment of the present disclosure, the deep learning model includes a first feature extraction module, a second feature extraction module, and an association module.

According to an embodiment of the present disclosure, the above method further includes:

processing the sample historical browsing track feature sequence and the sample user profile feature sequence to obtain second sample recommendation information corresponding to the sample user; and

and training the first feature extraction module and the second feature extraction module by using the second sample recommendation information and the real recommendation information to obtain a trained first feature extraction module and a trained second feature extraction module.

According to an embodiment of the present disclosure, training the deep learning model using the first sample recommendation information and the real recommendation information to obtain an information recommendation model includes:

and training the association module by using the first sample recommendation information and the real recommendation information under the condition that model parameters of the trained first feature extraction module and the trained second feature extraction module are kept unchanged, so as to obtain the information recommendation model.

According to an embodiment of the disclosure, the first feature extraction module includes a converter-based bi-directional encoding model, the second feature extraction module includes a factorizer model, and the correlation module includes a deep neural network model.

According to another aspect of the present disclosure, there is provided an information recommendation apparatus including:

the first acquisition module is used for responding to the detection of the operation behavior of a user aiming at a target link and acquiring user information corresponding to the user, wherein the user information comprises a historical browsing track information set and a user profile information set;

the first processing module is used for processing the historical browsing track information set to obtain a historical browsing track characteristic sequence;

the second processing module is used for processing the user profile information set to obtain a user profile feature sequence; and

and the first determining module is used for determining recommendation information corresponding to the user according to the historical browsing track feature sequence and the user profile feature sequence.

According to another aspect of the present disclosure, there is provided a training apparatus of an information recommendation model, including:

the second acquisition module is used for acquiring sample user information corresponding to the sample user, wherein the sample user information comprises a sample historical browsing track information set, a sample user profile information set and real recommendation information;

the third processing module is used for processing the sample history browsing track information set to obtain a sample history browsing track characteristic sequence;

The fourth processing module is used for processing the sample user profile information set to obtain a sample user profile feature sequence;

the second determining module is used for determining first sample recommendation information corresponding to the sample user according to the sample historical browsing track feature sequence and the sample user profile feature sequence; and

and the first training module is used for training the deep learning model by using the first sample recommendation information and the real recommendation information to obtain an information recommendation model.

According to another aspect of the present disclosure, there is provided an electronic device including:

one or more processors;

a memory for storing one or more instructions,

wherein the one or more instructions, when executed by the one or more processors, cause the one or more processors to implement a method as described in the present disclosure.

According to another aspect of the present disclosure, there is provided a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to implement a method as described in the present disclosure.

According to another aspect of the present disclosure, there is provided a computer program product comprising computer executable instructions which, when executed, are adapted to carry out the method as described in the present disclosure.

According to the embodiment of the present disclosure, since the history browsing trajectory feature sequence is obtained by processing the history browsing trajectory information set, the user profile feature sequence is obtained by processing the user profile information set, and the history browsing trajectory information set and the user profile information set are obtained in response to detecting the operation behavior of the user with respect to the target link, the history browsing trajectory feature sequence and the user profile feature sequence can be used to characterize the feature information related to the user. On the basis, the technical problem that the recommending method in the related art lacks learning ability of text information is at least partially overcome by determining recommending information corresponding to a user according to the historical browsing track characteristic sequence and the user profile characteristic sequence, and the accuracy and the interpretability of the recommending information are improved because the recommending information is obtained based on the user information acquired after the operation behavior of the user for the target link is detected.

Drawings

The above and other objects, features and advantages of the present disclosure will become more apparent from the following description of embodiments thereof with reference to the accompanying drawings in which:

FIG. 1 schematically illustrates a system architecture to which a training method of an information recommendation method, an information recommendation model, may be applied, according to an embodiment of the present disclosure;

FIG. 2 schematically illustrates a flow chart of an information recommendation method according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates an example schematic diagram of a process for processing a set of historical browsing trajectory information to arrive at a sequence of historical browsing trajectory features, according to an embodiment of the present disclosure;

FIG. 4 schematically illustrates an example schematic diagram of a process for processing a set of user profile information to obtain a sequence of user profile features, according to an embodiment of the disclosure;

FIG. 5 schematically illustrates an example schematic diagram of a process of determining recommendation information corresponding to a user based on a historical browsing trajectory feature sequence and a user profile feature sequence according to an embodiment of the present disclosure;

FIG. 6 schematically illustrates a flow chart of a training method of an information recommendation model, according to an embodiment of the present disclosure;

FIG. 7 schematically illustrates an example schematic diagram of a training process of an information recommendation model, according to an embodiment of the disclosure;

FIG. 8 schematically illustrates a block diagram of an information recommendation device according to an embodiment of the present disclosure;

FIG. 9 schematically illustrates a block diagram of a training apparatus of an information recommendation model, according to an embodiment of the present disclosure; and

Fig. 10 schematically illustrates a block diagram of an electronic device adapted to implement an information recommendation method, a training method of an information recommendation model, according to an embodiment of the disclosure.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

In embodiments of the present disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc., of the data involved (including, but not limited to, user personal information) all comply with relevant legal regulations, are used for legal purposes, and do not violate well-known. In particular, necessary measures are taken for personal information of the user, illegal access to personal information data of the user is prevented, and personal information security, network security and national security of the user are maintained.

In embodiments of the present disclosure, the user's authorization or consent is obtained before the user's personal information is obtained or collected.

For example, after obtaining the user information, you's information may be desensitized in a manner that includes de-identification or anonymization to secure your information.

In the related art, the recommendation method generally includes one of the following ways: recommendation methods based on collaborative filtering and recommendation methods based on the Wide & Deep model.

Collaborative filtering-based recommendation methods may refer to recommending questions that may be of interest to a user based on similarities between users or between questions and users. The recommendation method based on the Wide & Deep model may refer to making recommendation by learning various information and correlations between behaviors and questions of the user using feature learning capabilities of the model.

However, since the above methods lack learning ability for text information, the relevance between text information cannot be fully utilized, resulting in failure to ensure accuracy of the recommendation problem and lack of interpretability.

In order to at least partially solve the technical problems in the related art, the disclosure provides an information recommendation method, a training method of an information recommendation model, a training device of the information recommendation model and training equipment of the information recommendation model, and the training method and the training equipment can be applied to the technical fields of artificial intelligence, deep learning, natural language processing and information recommendation. The information recommendation method comprises the following steps: acquiring user information of a user in response to detecting an operation behavior of the user aiming at a target link, wherein the user information comprises a historical browsing track information set and a user profile information set; processing the historical browsing track information set to obtain a historical browsing track feature sequence; processing the user profile information set to obtain a user profile feature sequence; and determining recommendation information corresponding to the user according to the historical browsing track feature sequence and the user profile feature sequence.

Fig. 1 schematically illustrates a system architecture to which a training method of an information recommendation method, an information recommendation model, may be applied according to an embodiment of the present disclosure. It should be noted that fig. 1 is only an example of a system architecture to which embodiments of the present disclosure may be applied to assist those skilled in the art in understanding the technical content of the present disclosure, but does not mean that embodiments of the present disclosure may not be used in other devices, systems, environments, or scenarios.

As shown in fig. 1, a system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is a medium used to provide a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user may interact with the server 105 through the network 104 using at least one of the first terminal device 101, the second terminal device 102, the third terminal device 103, to receive or send messages, etc. Various communication client applications, such as a shopping class application, a web browser application, a search class application, an instant messaging tool, a mailbox client, social platform software, etc. (by way of example only) may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103.

The first terminal device 101, the second terminal device 102, the third terminal device 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop and desktop computers, and the like.

The server 105 may be a server providing various services, such as a background management server (by way of example only) providing support for websites browsed by the user using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as the user request, and feed back the processing result (e.g., the web page, information, or data obtained or generated according to the user request) to the terminal device.

It should be noted that, the information recommendation method provided by the embodiment of the present disclosure may be generally performed by the first terminal device 101, the second terminal device 102, or the third terminal device 103, or may be performed by other terminal devices different from the first terminal device 101, the second terminal device 102, or the third terminal device 103. Accordingly, the information recommendation apparatus provided by the embodiments of the present disclosure may also be provided in the first terminal device 101, the second terminal device 102, or the third terminal device 103, or in other terminal devices different from the first terminal device 101, the second terminal device 102, or the third terminal device 103.

Alternatively, the information recommendation method provided by the embodiments of the present disclosure may also be performed by the server 105. Accordingly, the information recommendation apparatus provided by the embodiments of the present disclosure may be generally provided in the server 105. The information recommendation method provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and/or the server 105. Accordingly, the information recommendation apparatus provided by the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and/or the server 105.

It should be noted that, the training method of the information recommendation model provided in the embodiments of the present disclosure may be generally performed by the server 105. Accordingly, the training device of the information recommendation model provided in the embodiments of the present disclosure may be generally disposed in the server 105. The training method of the information recommendation model provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and/or the server 105. Accordingly, the training apparatus of the information recommendation model provided in the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and/or the server 105.

Alternatively, the training method of the information recommendation model provided by the embodiment of the present disclosure may also be performed by the first terminal device 101, the second terminal device 102, or the third terminal device 103, or may also be performed by other terminal devices different from the first terminal device 101, the second terminal device 102, or the third terminal device 103. Accordingly, the training apparatus of the information recommendation model provided in the embodiments of the present disclosure may also be provided in the first terminal device 101, the second terminal device 102, or the third terminal device 103, or in other terminal devices different from the first terminal device 101, the second terminal device 102, or the third terminal device 103.

It should be understood that the number of first, second or third terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of first, second or third terminal devices, networks and servers, as desired for implementation.

It should be noted that the sequence numbers of the respective operations in the following methods are merely representative of the operations for the purpose of description, and should not be construed as representing the order of execution of the respective operations. The method need not be performed in the exact order shown unless explicitly stated.

Fig. 2 schematically illustrates a flowchart of an information recommendation method according to an embodiment of the present disclosure.

As shown in fig. 2, the information recommendation method 200 includes operations S210 to S240.

In response to detecting the operational behavior of the user with respect to the target link, user information of the user is obtained in operation S210, wherein the user information includes a set of historical browsing track information and a set of user profile information.

In operation S220, the historical browsing track information set is processed to obtain a historical browsing track feature sequence.

In operation S230, the user profile information set is processed to obtain a user profile feature sequence.

In operation S240, recommendation information corresponding to the user is determined according to the history browsing trajectory feature sequence and the user profile feature sequence.

According to the embodiment of the disclosure, the information recommendation method can be applied to a problem recommendation scene of financial customer service. The target link may refer to a connection relationship from target a to target B. Target a and target B may include at least one of: different web pages, different locations on the same web page, pictures, email addresses, files or applications, etc. The operational behavior for the target link may include at least one of: click, hover, voice control, etc. The user may access, via object a, object B whose backend is associated with the object link based on the operational behavior performed with respect to the object link.

According to embodiments of the present disclosure, after detecting an operational behavior of a user with respect to a target link, a user identification corresponding to the user may be determined. After obtaining the user identification, a set of historical browsing track information and a set of user profile information may be obtained from the data source based on the user identification. The data source may include at least one of: local databases, cloud databases, and network resources. For example, a data interface may be invoked, with which a set of historical browsing trail information and a set of user profile information are obtained from a data source.

According to embodiments of the present disclosure, user information may refer to feature information associated with a user. The user information may include a set of historical browsing track information and a set of user profile information. The set of historical browsing trajectory information may include association relationship information between at least one historical browsing trajectory information and at least one historical browsing trajectory information. The historical browsing track information may be used to characterize the browsing track of the user. For example, the historical browsing track information may include "white-bar payouts" and "gold-bar borrowing" pages. The set of user profile information may include a plurality of user profile information. The user profile information may be used to characterize a personal profile of the user. For example, the user profile information may include at least one of: gender, age, arrears, overdue, orders, coupons, and the like.

According to the embodiment of the disclosure, after the set of historical browsing track information is obtained, the at least one historical browsing track information may be processed by using the first predetermined model to obtain the historical browsing track feature vectors corresponding to the at least one historical browsing track information respectively. On the basis, a historical browsing track feature sequence can be determined according to at least one historical browsing track feature vector.

According to the embodiment of the present disclosure, the first predetermined model may be configured according to actual service requirements, and may be capable of implementing an encoding function, which is not limited herein. For example, the first predetermined model may include at least one of: DSSM (Deep Structured Semantic Models, semantic model based on depth network), DRMM (Deep Relevance Matching Model, depth correlation matching model), match Pyramid (construct matching matrix), and BERT (Bidirectional Encoder Representation from Transformers, bi-directional encoder representation).

According to an embodiment of the present disclosure, after the user profile information set is obtained, the at least one user profile information may be processed using the second predetermined model to obtain user profile feature vectors corresponding to the at least one user profile information, respectively. On the basis of this, a user profile feature sequence can be determined from the at least one user profile feature vector.

According to the embodiment of the present disclosure, the second predetermined model may be configured according to actual service requirements, and may be capable of implementing a dense vector representation function of mapping sparse features of the high-dimensional space to the low-dimensional space, which is not limited herein. For example, the second predetermined model may include at least one of: self-encoder (i.e., autoencoder), convolutional neural network (Convolutional Neural Network, CNN), recurrent neural network (Recurrent Neural Network, RNN), and converter model (i.e., transducer).

According to an embodiment of the present disclosure, the historical browsing trajectory feature sequence may include at least one historical browsing trajectory feature vector. The user profile feature sequence may comprise at least one user profile feature vector. After the historical browsing track feature sequence and the user profile feature sequence are obtained, each historical browsing track feature vector in the at least one historical browsing track feature vector and each user profile feature vector in the at least one user profile feature vector can be respectively processed to obtain at least one intermediate feature vector. On the basis of this, recommendation information can be determined from the at least one intermediate feature vector.

According to the embodiment of the disclosure, after the recommendation information is obtained, the recommendation information can be displayed in the output area of the target page so as to be convenient for information recommendation to the user. The recommendation information may include commodity names, service names, commodity scores, user ratings, popular leaderboards, problem lists, and the like.

An information recommendation method 200 according to an embodiment of the present invention is further described below with reference to fig. 2 to 5.

According to an embodiment of the present disclosure, operation S210 may include the following operations.

In response to detecting an operational behavior of the user with respect to the target link, a user identification corresponding to the user is determined. And acquiring a historical browsing track information set and a user profile information set from the data source according to the user identification.

According to embodiments of the present disclosure, user identification may be used to characterize different users. The user identification may include at least one of: a uniform resource identifier (Uniform Resource Identifier, URI) and an identification number (Identity Document, ID). The uniform resource identifier may include a uniform resource location system (Uniform Resource Locator, URL) and a uniform resource name (Uniform Resource Name, URN).

According to embodiments of the present disclosure, after obtaining the user identification, a set of historical browsing track information and a set of user profile information may be obtained from the user identification. The historical browsing track information sets and the user profile information sets may be obtained by real-time collection. For example, the information can be acquired by acquiring input information and click information of a user. Alternatively, the set of historical browsing track information and the set of user profile information may be obtained from a data source. For example, the data source may include at least one of: local databases, cloud databases, and network resources. Alternatively, the set of historical browsing track information and the set of user profile information may be received from other terminal devices. The embodiment of the disclosure does not limit the acquisition mode of the historical browsing track information set and the user profile information set.

According to an embodiment of the present disclosure, operation S220 may include the following operations.

And according to the association relation information, encoding at least one piece of historical browsing track information to obtain at least one position code and a historical browsing track feature vector corresponding to the at least one position code. And determining a historical browsing track feature sequence according to the at least one position code and the historical browsing track feature vector corresponding to the at least one position code.

According to an embodiment of the present disclosure, the set of historical browsing trajectory information may include association relationship information between at least one historical browsing trajectory information and at least one historical browsing trajectory information.

According to embodiments of the present disclosure, the association information may be used to characterize an association of at least one historical browsing trajectory information with each other. The process of processing the historical browsing track information set to obtain the historical browsing track feature sequence can be shown in the following formulas (1) and (2).

X _text ＝[x ₁ ，x ₂ ，x ₃ ，...，x _t ] (1)

S _text ＝Sum(BERT(X _text ))＝Sum([s ₁ ，s ₂ ，s ₃ ，...，s _t ]) (2)

Wherein X is _text Characterizing a historical browsing trajectory information set, e.g., x ₁ Characterization of "white strip payoff Page", x ₂ Characterization "gold bar borrowing page", x ₃ Characterization "Mobile phone recharge Page", …, x _t Characterization of "XX Page". S is S _text Characterizing a sequence of historical browsing trajectory features, e.g. S _text Is X _text Output sequence s after BERT coding ₁ Characterizing a historical browsing trajectory feature vector, s, corresponding to position code 1 ₂ Characterizing a historical browsing trajectory feature vector, s, corresponding to position code 2 ₃ Characterizing historical browsing trajectory feature vectors corresponding to position codes 3, …, s _t And characterizing the historical browsing track feature vector corresponding to the position code t.

According to the embodiment of the disclosure, the historical browsing track feature vectors corresponding to the at least one position code are obtained by coding the at least one historical browsing track information, so that interests of a user can be associated with and understood from historical behaviors, and subsequent recommendation accuracy is improved. On the basis, the historical browsing track feature sequence is determined according to at least one position code and the historical browsing track feature vector corresponding to the at least one position code, so that the historical browsing track information set can be converted into useful information, and the subsequent recommendation accuracy is further improved.

Fig. 3 schematically illustrates an example schematic diagram of a process for processing a set of historical browsing trajectory information to obtain a sequence of historical browsing trajectory features according to an embodiment of the present disclosure.

As shown in fig. 3, in 300, a set of historical browsing trajectory information 301 includes at least one historical browsing trajectory information and association relationship information between the at least one historical browsing trajectory information. For example, the at least one historical browsing trace information may include historical browsing trace information 301_1, historical browsing trace information 301_2, …, historical browsing trace information 301_x, …, historical browsing trace information 301X. X may be an integer greater than or equal to 1, X ε {1,2, …, (X-1), X }.

And respectively encoding at least one historical browsing track information according to the association relation information to obtain at least one position code and a historical browsing track feature vector corresponding to the at least one position code. For example, the history browsing trajectory information 301_1 may be subjected to encoding processing, resulting in the position code 302_1 and the history browsing trajectory feature vector 303_1 corresponding to the position code 302_1. The history browsing trajectory information 301_2 is subjected to encoding processing, and a position code 302_2 and a history browsing trajectory feature vector 303_2 corresponding to the position code 302_2 are obtained.

Similarly, the history browsing track information 301_x may be encoded to obtain the position code 302_x and the history browsing track feature vector 303_x corresponding to the position code 302_x. By analogy, the history browsing trajectory information 301X is subjected to encoding processing, and a position code 302_x and a history browsing trajectory feature vector 303_x corresponding to the position code 302_x are obtained.

On this basis, the historical browsing trajectory feature sequence 304 may be determined from the position codes 302_1, 302_2, …, 302_x, …, 302_x, and the historical browsing trajectory feature vector 3031, the historical browsing trajectory feature vectors 303_2, …, the historical browsing trajectory feature vectors 303_x, …, and the historical browsing trajectory feature vector 303_x.

According to an embodiment of the present disclosure, operation S230 may include the following operations.

For user profile information of the plurality of user profile information, a profile information type of the user profile information is determined based on the user profile information. In the case where the profile information type is a continuous value type, a predetermined continuous encoding mode corresponding to the continuous value type is determined. The user profile information is encoded based on a predetermined continuous encoding scheme to obtain a first user profile feature corresponding to the user profile information.

According to embodiments of the present disclosure, the set of user profile information may include a plurality of user profile information.

According to an embodiment of the present disclosure, the profile information type may include one of: a continuous numeric type and a discrete category type. A continuous numerical type may refer to data that can be represented by real numbers, typically any value within a certain range of values. Such as height and weight, etc. In processing data of continuous numeric types, modeling and prediction can be performed using statistical methods and machine learning algorithms. Discrete category types may refer to data, typically some sort of classification or label, that needs to be represented with a limited number of values. Such as gender occupation, etc. In processing discrete class type data, it is often necessary to convert it into a numerical representation and use corresponding statistical methods and machine learning algorithms for modeling and prediction.

According to an embodiment of the present disclosure, the process of processing a user profile information set to obtain a user profile feature sequence may be represented by the following formulas (3) and (4).

Y _user ＝[y ₁ ，y ₂ ，y ₃ ，...，y _k ] (3)

H _tag ＝[h ₁ ，h ₂ ，h ₃ ，...，h _k ] (4)

Wherein Y is _user Characterizing a set of user profile information, e.g., y ₁ Characterization of sex, y ₂ Characterising age, y ₃ Characterization of overdue, …, y _k The order is characterized. H _tag Characterizing user profile feature sequences, e.g. H _tag Is Y _user Vector after being subjected to Embedding, h ₁ Characterizing a user profile feature vector, h, corresponding to position code 1 ₂ Characterizing a user profile feature vector, h, corresponding to position code 2 ₃ Characterizing user profile feature vectors corresponding to position codes 3, …, h _k The user profile feature vector corresponding to the position code k is characterized.

According to embodiments of the present disclosure, a predetermined continuous encoding manner may be determined according to a specific category, a specific scenario, and a task of user profile information. The predetermined consecutive encoding mode corresponding to the consecutive numerical value type may include at least one of: discretization (i.e., discounting), normalization (i.e., normalization), binning (i.e., binning), logarithmic transformation (i.e., logarithmic Transformation), polynomial features (i.e., polynomial Feature), and Basis functions (i.e., basis Function). On this basis, the user profile information may be encoded based on a predetermined continuous encoding scheme, resulting in a first user profile feature corresponding to the user profile information.

According to an embodiment of the present disclosure, the above-described method may further include the following operations.

In the case where the profile information type is a discrete category type, a predetermined discrete coding scheme corresponding to the discrete category type is determined. And encoding the user profile information based on a predetermined discrete encoding mode to obtain a second user profile characteristic corresponding to the user profile information.

According to embodiments of the present disclosure, the predetermined discrete coding manner may be determined according to a specific category, a specific scenario, and a task of the user profile information. The predetermined discrete coding pattern corresponding to the discrete category type may include at least one of: one-Hot Encoding (i.e., one-Hot Encoding), tag Encoding (i.e., label Encoding), and Binary Encoding (i.e., binary Encoding). On this basis, the user profile information may be encoded based on a predetermined discrete encoding scheme to obtain a second user profile feature corresponding to the user profile information.

According to the embodiment of the disclosure, the profile information type is determined according to the user profile information, so that the predetermined coding mode is determined according to different profile information types, and the processing efficiency of data is improved. On the basis, the first user profile characteristic is obtained by encoding the user profile information based on a preset continuous encoding mode, the second user profile characteristic is obtained by encoding the user profile information based on a preset discrete encoding mode, and the profile information of the user can be more accurately represented through the first user profile characteristic and the second user profile characteristic, so that the accuracy of data processing is improved, and the accuracy of follow-up information recommendation is further improved.

FIG. 4 schematically illustrates an example schematic diagram of a process for processing a set of user profile information to obtain a sequence of user profile features, according to an embodiment of the disclosure.

As shown in fig. 4, in 400, a profile information type 402 of the user profile information 401 may be determined from the user profile information 401. After obtaining the profile information type 402, operation S410 may be performed.

Is the profile information type a continuous numeric type?

If so, a predetermined continuous coding scheme 403 corresponding to the continuous value type may be determined, and the user profile information 401 may be coded based on the predetermined continuous coding scheme 403, so as to obtain a first user profile feature 404 corresponding to the user profile information 401.

If not, a predetermined discrete coding scheme 405 corresponding to the discrete category type may be determined, and the user profile information 401 may be coded based on the predetermined discrete coding scheme 405 to obtain the second user profile feature 406 corresponding to the user profile information 401.

According to an embodiment of the present disclosure, operation S240 may include the following operations.

And processing the historical browsing track feature sequence and the user profile feature sequence to obtain an intermediate feature vector set. And processing the intermediate feature vector set to obtain recommendation information.

According to an embodiment of the present disclosure, the historical browsing trajectory feature sequence and the user profile feature sequence may be processed using a third predetermined model to obtain an intermediate feature vector set. The specific processing manner may be configured according to the actual service requirement, which is not limited herein. For example, vector concatenation (i.e., vector Concatenation) processing, vector weighted average (i.e., weighted Vector Averaging) processing, lexical Analysis (i.e., lexical Analysis) processing, and vector feature intersection (i.e., vector Feature Crossing) processing.

According to the embodiment of the disclosure, after the intermediate feature vector set is obtained, the intermediate feature vector set may be processed by using a fourth predetermined model to obtain recommendation information. The fourth predetermined model may be configured according to actual service requirements, which is not limited herein. For example, the fourth predetermined model may include at least one of: convolutional neural networks, recurrent neural networks, deep neural networks (Deep Neural Network, DNN), long Short-Term Memory networks (LSTM), and generative antagonism networks (Generative Adversarial Network, GAN).

According to embodiments of the present disclosure, processing the historical browsing trajectory feature sequence and the user profile feature sequence to obtain the set of intermediate feature vectors may include the following operations.

According to an embodiment of the present disclosure, processing the intermediate feature vector set to obtain the recommendation information may include the following operations.

And respectively carrying out fusion processing on the explicit feature vector and the Q implicit feature vectors aiming at each explicit feature vector in the P explicit feature vectors to obtain Q target feature vectors. And determining recommendation information according to the Q target feature vectors corresponding to the P explicit feature vectors.

According to an embodiment of the present disclosure, the historical browsing trajectory feature sequence may include M historical browsing trajectory feature vectors. The user profile feature sequence may include N user profile feature vectors. M and N are positive integers.

In accordance with embodiments of the present disclosure, feature interleaving may refer to the operation of bit-wise multiplying, adding, etc., two or more vectors to produce a new feature vector. The feature cross processing can better express the interaction information in the original features, and is beneficial to improving the performance of the machine learning model. For example, for a given user and commodity, their attribute vectors may be multiplied or added bit by bit, with the resulting new feature vector representing interaction information between them, including user preferences for the commodity, attribute features of the commodity, and so on. Alternatively, the word vector and the text position vector may be multiplied or added bit by bit, and the interaction of the position information and the semantic information of the word in the text is represented by the resulting new feature vector.

According to embodiments of the present disclosure, the fusion process may refer to combining a plurality of different feature vectors to form a new, more representative feature vector. The feature vector fusion process may include feature extraction, feature selection, feature dimension adjustment, and feature vector fusion. The fusion processing manner may be configured according to actual service requirements, which is not limited herein. For example, the fusion process may include at least one of: splicing (i.e., localization) fusion, weighted (i.e., weighted) fusion, stacked generalized (i.e., stacking) fusion, hybrid integration (i.e., blending) fusion, and boost integration (i.e., boosting) fusion.

According to an embodiment of the present disclosure, performing feature intersection processing on M historical browsing trajectory feature vectors and N user profile feature vectors, obtaining P explicit feature vectors and Q implicit feature vectors may include the following operations.

And aiming at each historical browsing track feature vector in the M historical browsing track feature vectors, aiming at each user profile feature vector in the N user profile feature vectors, carrying out low-order feature cross processing on the historical browsing track feature vectors and the user profile feature vectors to obtain an explicit feature vector. And performing high-order feature cross processing on the historical browsing track feature vector and the user profile feature vector to obtain an implicit feature vector.

According to the embodiment of the disclosure, the explicit feature vector can be obtained by performing low-order feature cross processing on the historical browsing track feature vector and the user profile feature vector. The explicit feature vector may refer to a feature that is manually designed or selected, and may generally include basic features such as numerical values, categories, and the like in the original data, and may also perform feature construction and selection according to domain knowledge and a specific task.

In accordance with an embodiment of the present disclosure, low-order feature intersection processing may refer to an intersection between two or three features, e.g., an intersection of feature a and feature B may result in a new feature AB. Low-order feature interleaving may be used to process features that belong to a continuous numerical type. For example, crossing the two features of Height (i.e., height) and Weight (i.e., weight) may yield new features of BMI (i.e., body Mass Index).

According to the embodiment of the disclosure, the implicit feature vector can be obtained by performing high-order feature cross processing on the historical browsing track feature vector and the user profile feature vector. Implicit feature vectors may refer to feature vectors that are automatically extracted from the raw data by a machine learning algorithm. Implicit feature vectors are typically high-dimensional with strong expressive power.

In accordance with embodiments of the present disclosure, high-order feature intersection processing may refer to intersections between more features, e.g., intersections between four or more features. High-order feature interleaving can be used to process features belonging to discrete class types, enabling better capture of nonlinear relationships and complex structures of data. For example, new features of the user representation are obtained according to a combination of features such as Gender (i.e., gender), age (i.e., age), occupation (i.e., location), and region (i.e., location).

According to the embodiment of the disclosure, the explicit feature vector is obtained by performing low-order feature cross processing on the historical browsing track feature vector and the user profile feature vector, and the implicit feature vector is obtained by performing high-order feature cross processing on the historical browsing track feature vector and the user profile feature vector, so that the explicit feature vector and the implicit feature vector can better reflect the historical browsing track feature and the user summary feature, and the efficiency and the accuracy of the follow-up information recommendation are improved. On the basis, the target feature vector is obtained by fusion processing of the explicit feature vector and the implicit feature vector, so that the problem recommendation service based on the personalized requirements of the user can be realized, and the satisfaction degree of the user can be improved.

Fig. 5 schematically illustrates an example schematic diagram of a process of determining recommendation information corresponding to a user according to a historical browsing trajectory feature sequence and a user profile feature sequence according to an embodiment of the present disclosure.

As shown in fig. 5, at 500, a low-level feature cross process may be performed on the historical browsing trajectory feature vector 501 and the user profile feature vector 502 to obtain an explicit feature vector 503. The historical browsing trajectory feature vector 501 and the user profile feature vector 502 are subjected to high-order feature cross processing to obtain an implicit feature vector 504.

On this basis, fusion processing can be performed on the explicit feature vector 503 and Q implicit feature vectors 504, respectively, to obtain Q target feature vectors 505. Recommendation information 506 is determined based on Q target feature vectors 505 corresponding to each of the P explicit feature vectors 503.

According to an embodiment of the present disclosure, the information recommendation method 200 may further include the following operations.

And displaying the recommendation information in the output area of the target page.

According to embodiments of the present disclosure, a target page may refer to a page associated with a target link. The output area may include a target location and/or a target interface. After the recommendation information is obtained, the recommendation information can be displayed to the user at the target position and/or the target interface of the target page.

For example, the recommendation information may be "how overdue the white strip is" and the recommendation corresponding to the recommendation information is "overdue the white strip account of the user". Alternatively, the recommendation information may be "how the funds are reddish", and the recommendation reason corresponding to the recommendation information is "the user frequently browses the fund reddish pages".

The above is only an exemplary embodiment, but is not limited thereto, and other information recommendation methods known in the art may be included as long as the accuracy and interpretability of the recommended information can be improved.

Fig. 6 schematically illustrates a flowchart of a training method of an information recommendation model according to an embodiment of the present disclosure.

As shown in fig. 6, the training method 600 of the information recommendation model includes operations S610 to S650.

Sample user information corresponding to the sample user is acquired in operation S610, wherein the sample user information includes a sample history browsing track information set, a sample user profile information set, and real recommendation information.

In operation S620, the sample history browsing track information set is processed to obtain a sample history browsing track feature sequence.

In operation S630, the sample user profile information set is processed to obtain a sample user profile feature sequence.

In operation S640, first sample recommendation information corresponding to the sample user is determined according to the sample history browsing trajectory feature sequence and the sample user profile feature sequence.

In operation S650, the deep learning model is trained using the first sample recommendation information and the real recommendation information to obtain an information recommendation model.

According to embodiments of the present disclosure, the actual recommendation information may refer to actual recommendation information corresponding to a sample user. For the description of the sample user information, the sample historical browsing track information set, the sample user profile information set, the sample historical browsing track feature sequence, the sample user profile feature sequence and the first sample recommendation information, reference may be made to the relevant content of the user information, the historical browsing track information set, the user profile information set, the historical browsing track feature sequence, the user profile feature sequence and the recommendation information, which are not described herein.

According to the embodiment of the disclosure, since the sample history browsing track feature sequence is obtained by processing the sample history browsing track information set, and the sample user profile feature sequence is obtained by processing the sample user profile information set, the sample history browsing track information set and the sample user profile information set are corresponding to the sample user, the sample history browsing track feature sequence and the sample user profile feature sequence can be used for characterizing feature information related to the sample user. On the basis, the first sample recommendation information corresponding to the sample user is determined according to the sample historical browsing track feature sequence and the sample user profile feature sequence, and the deep learning model is trained by utilizing the first sample recommendation information and the real recommendation information.

A training method 600 of the information recommendation model according to an embodiment of the present invention is further described below with reference to fig. 7.

According to an embodiment of the present disclosure, the training method 600 of the information recommendation model may further include the following operations.

And processing the sample historical browsing track feature sequence and the sample user profile feature sequence to obtain second sample recommendation information corresponding to the sample user. And training the first feature extraction module and the second feature extraction module by using the second sample recommendation information and the real recommendation information to obtain a trained first feature extraction module and a trained second feature extraction module.

According to an embodiment of the present disclosure, operation S650 may include the following operations.

And under the condition that model parameters of the trained first feature extraction module and the trained second feature extraction module are kept unchanged, training the association module by using the first sample recommendation information and the real recommendation information to obtain an information recommendation model.

According to an embodiment of the present disclosure, a deep learning model may include a first feature extraction module, a second feature extraction module, and an association module.

According to an embodiment of the present disclosure, the first feature extraction module comprises a converter-based bi-directional encoding model, the second feature extraction module comprises a factorizer model, and the correlation module comprises a deep neural network model.

According to embodiments of the present disclosure, the converter-based bi-directional coding model is adapted to process textual information, and the factorizer model and the deep neural network model are adapted to process numerical information. The problems of interest of the user can be mined according to the user click log and the user consultation log of the financial customer service scene, and the problems of interest of the user are set as positive samples. The exposed and not clicked problem can be set as a negative sample.

According to the embodiment of the disclosure, the first loss function value may be obtained based on the first loss function by using the real recommendation information and the second sample recommendation information. And adjusting model parameters of the first feature extraction module and the second feature extraction module according to the first loss function value until a preset condition is met. For example, the model parameters of the first feature extraction module and the second feature extraction module may be adjusted according to a back propagation algorithm or a random gradient descent algorithm until a predetermined condition is met. The first feature extraction module and the second feature extraction module obtained in the case that the predetermined condition is satisfied are determined as the trained first feature extraction module and the trained second feature extraction module.

According to the embodiment of the disclosure, the second loss function value may be obtained based on the second loss function by using the real classification result and the first sample recommendation information. And adjusting the model parameters of the association module according to the second loss function value until a preset condition is met. For example, the model parameters of the associated modules may be adjusted according to a back-propagation algorithm or a random gradient descent algorithm until a predetermined condition is met. The trained first feature extraction module, the trained second feature extraction module, and the association module obtained if the predetermined condition is satisfied are determined as an information recommendation model.

According to an embodiment of the present disclosure, the first and second loss functions may comprise, for example, a range loss function, an exponential loss function, a square loss function, or a cross entropy loss function. The predetermined condition may include at least one of convergence of the output value and reaching of the training round to a maximum training round.

According to the embodiment of the disclosure, the BERT model is utilized to learn the characteristic representation of the sample historical browsing track information set by combining the BERT model and the deep FM model, and the characteristic representation of the sample historical browsing track information set and the sample user profile information set are used as the characteristic input of the deep FM model, so that the recommending capability of the information recommending model is improved, and the accuracy and the interpretability of recommending learning are improved.

FIG. 7 schematically illustrates an example schematic diagram of a training process of an information recommendation model, according to an embodiment of the disclosure.

As shown in fig. 7, in 700, a deep learning model 702 may include a first feature extraction module 702_1, a second feature extraction module 702_2, and an association module 702_3.

Sample user information 701 corresponding to a sample user may be acquired, the sample user information 701 including a sample history browsing trajectory information set 701_1, a sample user profile information set 701_2, and real recommendation information 701_3.

The sample history browsing track information set 701_1 is input to the first feature extraction module 702_1, so as to obtain a sample history browsing track feature sequence 704. The sample user profile information set 701_2 is input to a second feature extraction module 702_2 resulting in a sample user profile feature sequence 705. The sample historical browsing trajectory feature sequence 704 and the sample user profile feature sequence 705 are input to the association module 702_3, resulting in first sample recommendation information 706.

Based on the loss function 707, a loss function value 708 is obtained using the real recommendation information 701_3 and the first sample recommendation information 706. Model parameters of the deep learning model 702 are adjusted according to the loss function values 708 until a predetermined condition is satisfied. The deep learning model 702 obtained in the case where the predetermined condition is satisfied is determined as the information recommendation model.

The above is only an exemplary embodiment, but is not limited thereto, and other training methods of the information recommendation model known in the art may be included as long as the information recommendation capability of the model can be improved.

Fig. 8 schematically illustrates a block diagram of an information recommendation device according to an embodiment of the present disclosure.

As shown in fig. 8, the information recommendation apparatus 800 may include a first acquisition module 810, a first processing module 820, a second processing module 830, and a first determination module 840.

The first obtaining module 810 is configured to obtain, in response to detecting an operation behavior of a user with respect to a target link, user information corresponding to the user, where the user information includes a historical browsing track information set and a user profile information set.

The first processing module 820 is configured to process the historical browsing track information set to obtain a historical browsing track feature sequence.

And a second processing module 830, configured to process the user profile information set to obtain a user profile feature sequence.

The first determining module 840 is configured to determine recommendation information corresponding to the user according to the historical browsing track feature sequence and the user profile feature sequence.

According to an embodiment of the present disclosure, the first determination module 840 may include a first processing sub-module and a second processing sub-module.

And the first processing sub-module is used for processing the historical browsing track feature sequence and the user profile feature sequence to obtain an intermediate feature vector set.

And the second processing sub-module is used for processing the intermediate feature vector set to obtain recommendation information.

According to an embodiment of the present disclosure, the historical browsing trajectory feature sequence includes M historical browsing trajectory feature vectors, the user profile feature sequence includes N user profile feature vectors, and M and N are positive integers.

According to an embodiment of the present disclosure, the first processing sub-module may comprise a cross processing unit.

And the cross processing unit is used for carrying out feature cross processing on the M historical browsing track feature vectors and the N user profile feature vectors to obtain P explicit feature vectors and Q implicit feature vectors, wherein P and Q are positive integers.

According to an embodiment of the present disclosure, the second processing sub-module may include a fusion processing unit and a determining unit.

The fusion processing unit is used for respectively carrying out fusion processing on the explicit feature vector and the Q implicit feature vectors aiming at each explicit feature vector in the P explicit feature vectors to obtain Q target feature vectors.

And the determining unit is used for determining recommendation information according to the Q target feature vectors corresponding to the P explicit feature vectors.

According to an embodiment of the present disclosure, the interleaving unit may include a first interleaving subunit and a second interleaving subunit.

The first cross processing subunit is configured to perform low-order feature cross processing on the historical browsing track feature vector and the user profile feature vector for each historical browsing track feature vector in the M historical browsing track feature vectors, and for each user profile feature vector in the N user profile feature vectors, to obtain an explicit feature vector.

And the second cross processing subunit is used for performing high-order feature cross processing on the historical browsing track feature vector and the user profile feature vector to obtain an implicit feature vector.

According to an embodiment of the present disclosure, the set of user profile information includes a plurality of user profile information.

According to an embodiment of the present disclosure, the second processing module 830 may include a first determination submodule, a second determination submodule, and a first encoding submodule.

A first determining sub-module for determining a profile information type of the user profile information according to the user profile information for the user profile information of the plurality of user profile information.

And the second determining submodule is used for determining a preset continuous coding mode corresponding to the continuous numerical value type when the profile information type is the continuous numerical value type.

And the first coding sub-module is used for coding the user profile information based on a preset continuous coding mode to obtain a first user profile characteristic corresponding to the user profile information.

The second processing module 830 may also include a third determination submodule and a second encoding submodule according to embodiments of the present disclosure.

And the third determining submodule is used for determining a preset discrete coding mode corresponding to the discrete category type when the profile information type is the discrete category type.

And the second coding sub-module is used for coding the user profile information based on a preset discrete coding mode to obtain a second user profile characteristic corresponding to the user profile information.

According to an embodiment of the present disclosure, the first processing module 820 may include a third encoding sub-module and a fourth determining sub-module.

And the third coding sub-module is used for carrying out coding processing on at least one historical browsing track information according to the association relation information to obtain at least one position code and a historical browsing track feature vector corresponding to the at least one position code.

And the fourth determining submodule is used for determining a historical browsing track feature sequence according to the at least one position code and the historical browsing track feature vector corresponding to the at least one position code.

According to an embodiment of the present disclosure, the first acquisition module 810 may include a fifth determination sub-module and an acquisition sub-module.

And a fifth determining sub-module, configured to determine a user identifier corresponding to the user in response to detecting an operation behavior of the user with respect to the target link.

And the acquisition sub-module is used for acquiring the historical browsing track information set and the user profile information set from the data source according to the user identification.

According to an embodiment of the present disclosure, the information recommendation apparatus 800 may further include a presentation module.

And the display module is used for displaying the recommendation information in the output area of the target page.

Fig. 9 schematically illustrates a block diagram of a training apparatus of an information recommendation model according to an embodiment of the present disclosure.

As shown in fig. 9, the training apparatus 900 of the information recommendation model may include a second acquisition module 910, a third processing module 920, a fourth processing module 930, a second determination module 940, and a first training module 950.

The second obtaining module 910 is configured to obtain sample user information corresponding to a sample user, where the sample user information includes a sample historical browsing track information set, a sample user profile information set, and real recommendation information.

And the third processing module 920 is configured to process the sample historical browsing track information set to obtain a sample historical browsing track feature sequence.

A fourth processing module 930 is configured to process the sample user profile information set to obtain a sample user profile feature sequence.

A second determining module 940, configured to determine first sample recommendation information corresponding to the sample user according to the sample history browsing track feature sequence and the sample user profile feature sequence.

The first training module 950 is configured to train the deep learning model to obtain an information recommendation model by using the first sample recommendation information and the real recommendation information.

According to an embodiment of the present disclosure, a deep learning model includes a first feature extraction module, a second feature extraction module, and an association module.

According to an embodiment of the present disclosure, the training apparatus 900 of the information recommendation model may further include a fifth processing module and a second training module.

And the fifth processing module is used for processing the sample history browsing track feature sequence and the sample user profile feature sequence to obtain second sample recommendation information corresponding to the sample user.

And the second training module is used for training the first feature extraction module and the second feature extraction module by using the second sample recommendation information and the real recommendation information to obtain a trained first feature extraction module and a trained second feature extraction module.

According to an embodiment of the present disclosure, the first training module 950 may include a training sub-module.

And the training sub-module is used for training the association module by using the first sample recommendation information and the real recommendation information under the condition of keeping the model parameters of the trained first feature extraction module and the trained second feature extraction module unchanged, so as to obtain an information recommendation model.

Any number of modules, sub-modules, units, sub-units, or at least some of the functionality of any number of the sub-units according to embodiments of the present disclosure may be implemented in one module. Any one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be implemented as split into multiple modules. Any one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be implemented at least in part as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-chip, a system-on-substrate, a system-on-package, an Application Specific Integrated Circuit (ASIC), or in any other reasonable manner of hardware or firmware that integrates or encapsulates the circuit, or in any one of or a suitable combination of three of software, hardware, and firmware. Alternatively, one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be at least partially implemented as computer program modules, which when executed, may perform the corresponding functions.

For example, any of the first acquisition module 810, the first processing module 820, the second processing module 830, and the first determination module 840 may be combined in one module/unit/sub-unit, or any of the modules/units/sub-units may be split into a plurality of modules/units/sub-units. Alternatively, at least some of the functionality of one or more of these modules/units/sub-units may be combined with at least some of the functionality of other modules/units/sub-units and implemented in one module/unit/sub-unit. According to embodiments of the present disclosure, at least one of the first acquisition module 810, the first processing module 820, the second processing module 830, and the first determination module 840 may be implemented at least in part as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented in hardware or firmware in any other reasonable manner of integrating or packaging the circuitry, or in any one of or a suitable combination of three of software, hardware, and firmware. Alternatively, at least one of the first acquisition module 810, the first processing module 820, the second processing module 830, and the first determination module 840 may be at least partially implemented as computer program modules, which when executed, may perform the respective functions.

It should be noted that, in the embodiment of the present disclosure, the information recommending apparatus portion corresponds to the information recommending method portion in the embodiment of the present disclosure, and the description of the information recommending apparatus portion specifically refers to the information recommending method portion and is not described herein again.

For example, any of the second acquisition module 910, the third processing module 920, the fourth processing module 930, the second determination module 940, and the first training module 950 may be combined in one module/unit/sub-unit, or any of the modules/units/sub-units may be split into a plurality of modules/units/sub-units. Alternatively, at least some of the functionality of one or more of these modules/units/sub-units may be combined with at least some of the functionality of other modules/units/sub-units and implemented in one module/unit/sub-unit. According to embodiments of the present disclosure, at least one of the second acquisition module 910, the third processing module 920, the fourth processing module 930, the second determination module 940, and the first training module 950 may be implemented at least in part as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or as hardware or firmware in any other reasonable manner of integrating or packaging the circuitry, or as any one of or a suitable combination of any of the three implementations of software, hardware, and firmware. Alternatively, at least one of the second acquisition module 910, the third processing module 920, the fourth processing module 930, the second determination module 940, and the first training module 950 may be at least partially implemented as computer program modules, which when executed, may perform the respective functions.

It should be noted that, in the embodiment of the present disclosure, the training device portion of the information recommendation model corresponds to the training method portion of the information recommendation model in the embodiment of the present disclosure, and the description of the training device portion of the information recommendation model specifically refers to the training method portion of the information recommendation model, which is not described herein.

Fig. 10 schematically illustrates a block diagram of an electronic device adapted to implement an information recommendation method, a training method of an information recommendation model, according to an embodiment of the disclosure. The electronic device shown in fig. 10 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.

As shown in fig. 10, a computer electronic device 1000 according to an embodiment of the present disclosure includes a processor 1001 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 1002 or a program loaded from a storage section 1009 into a Random Access Memory (RAM) 1003. The processor 1001 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or an associated chipset and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), or the like. The processor 1001 may also include on-board memory for caching purposes. The processor 1001 may include a single processing unit or multiple processing units for performing different actions of the method flows according to embodiments of the present disclosure.

In the RAM 1003, various programs and data necessary for the operation of the electronic apparatus 1000 are stored. The processor 1001, the ROM 1002, and the RAM 1003 are connected to each other by a bus 1004. The processor 1001 performs various operations of the method flow according to the embodiment of the present disclosure by executing programs in the ROM 1002 and/or the RAM 1003. Note that the program may be stored in one or more memories other than the ROM 1002 and the RAM 1003. The processor 1001 may also perform various operations of the method flow according to the embodiments of the present disclosure by executing programs stored in the one or more memories.

According to an embodiment of the disclosure, the electronic device 1000 may also include an input/output (I/O) interface 1005, the input/output (I/O) interface 1005 also being connected to the bus 1004. The electronic device 1000 may also include one or more of the following components connected to an input/output (I/O) interface 1005: an input section 1006 including a keyboard, a mouse, and the like; an output portion 1007 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), etc., and a speaker, etc.; a storage portion 1008 including a hard disk or the like; and a communication section 1009 including a network interface card such as a LAN card, a modem, or the like. The communication section 1009 performs communication processing via a network such as the internet. The drive 1010 is also connected to an input/output (I/O) interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like, is installed as needed in the drive 1010, so that a computer program read out therefrom is installed as needed in the storage section 1008.

According to embodiments of the present disclosure, the method flow according to embodiments of the present disclosure may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable storage medium, the computer program comprising program code for performing the method shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via the communication portion 1009, and/or installed from the removable medium 1011. The above-described functions defined in the system of the embodiments of the present disclosure are performed when the computer program is executed by the processor 1001. The systems, devices, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

The present disclosure also provides a computer-readable storage medium that may be embodied in the apparatus/device/system described in the above embodiments; or may exist alone without being assembled into the apparatus/device/system. The computer-readable storage medium carries one or more programs which, when executed, implement methods in accordance with embodiments of the present disclosure.

According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. Examples may include, but are not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

For example, according to embodiments of the present disclosure, the computer-readable storage medium may include ROM 1002 and/or RAM 1003 and/or one or more memories other than ROM 1002 and RAM 1003 described above.

Embodiments of the present disclosure also include a computer program product comprising a computer program comprising program code for performing the methods provided by the embodiments of the present disclosure, the program code for causing an electronic device to implement the information recommendation method, the training method of the information recommendation model provided by the embodiments of the present disclosure, when the computer program product is run on the electronic device.

The above-described functions defined in the system/apparatus of the embodiments of the present disclosure are performed when the computer program is executed by the processor 1001. The systems, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted in the form of signals on a network medium, distributed, and downloaded and installed via the communication section 1009, and/or installed from the removable medium 1011. The computer program may include program code that may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

According to embodiments of the present disclosure, program code for performing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, such computer programs may be implemented in high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. Programming languages include, but are not limited to, such as Java, c++, python, "C" or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions. Those skilled in the art will appreciate that the features recited in the various embodiments of the disclosure and/or in the claims may be combined in various combinations and/or combinations, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, the features recited in the various embodiments of the present disclosure and/or the claims may be variously combined and/or combined without departing from the spirit and teachings of the present disclosure. All such combinations and/or combinations fall within the scope of the present disclosure.

The embodiments of the present disclosure are described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be made by those skilled in the art without departing from the scope of the disclosure, and such alternatives and modifications are intended to fall within the scope of the disclosure.

Claims

1. An information recommendation method, comprising:

2. The method of claim 1, wherein the determining recommendation information corresponding to the user based on the historical browsing trajectory feature sequence and the user profile feature sequence comprises:

3. The method of claim 2, wherein the historical browsing trajectory feature sequence comprises M historical browsing trajectory feature vectors, the user profile feature sequence comprises N user profile feature vectors, M and N are positive integers;

and processing the historical browsing track feature sequence and the user profile feature sequence to obtain an intermediate feature vector set, wherein the processing comprises the following steps:

performing feature cross processing on the M historical browsing track feature vectors and the N user profile feature vectors to obtain P explicit feature vectors and Q implicit feature vectors, wherein P and Q are positive integers;

the processing the intermediate feature vector set to obtain the recommendation information includes:

for each of the P explicit feature vectors,

respectively carrying out fusion processing on the explicit feature vector and the Q implicit feature vectors to obtain Q target feature vectors; and

4. The method of claim 3, wherein the feature cross-processing the M historical browsing trajectory feature vectors and the N user profile feature vectors to obtain P explicit feature vectors and Q implicit feature vectors comprises:

for each of the M historical browsing trajectory feature vectors,

for each of the N user profile feature vectors,

performing low-order feature cross processing on the historical browsing track feature vector and the user profile feature vector to obtain the explicit feature vector; and

5. The method of any of claims 1 to 4, wherein the set of user profile information comprises a plurality of user profile information;

said processing said set of user profile information to obtain a user profile feature sequence comprising:

for a user profile of the plurality of user profile information,

Determining a profile information type of the user profile information according to the user profile information;

determining a predetermined continuous coding mode corresponding to the continuous value type when the profile information type is the continuous value type; and

and carrying out coding processing on the user profile information based on the preset continuous coding mode to obtain a first user profile characteristic corresponding to the user profile information.

6. The method of claim 5, further comprising:

determining a predetermined discrete coding mode corresponding to the discrete category type under the condition that the profile information type is the discrete category type; and

and carrying out coding processing on the user profile information based on the preset discrete coding mode to obtain a second user profile characteristic corresponding to the user profile information.

7. The method of any one of claims 1 to 4, wherein the set of historical browsing trajectory information includes at least one historical browsing trajectory information and association relationship information between the at least one historical browsing trajectory information;

the processing the historical browsing track information set to obtain a historical browsing track feature sequence comprises the following steps:

8. The method of any of claims 1-4, wherein the obtaining user information corresponding to a user in response to detecting an operational behavior of the user for a target link comprises:

9. The method of any of claims 1-4, further comprising, after the determining recommendation information corresponding to the user from the historical browsing trajectory feature sequence and the user profile feature sequence:

10. A training method of an information recommendation model, comprising:

processing the sample historical browsing track information set to obtain a sample historical browsing track feature sequence;

11. The method of claim 10, wherein the deep learning model comprises a first feature extraction module, a second feature extraction module, and an association module;

the method further comprises the steps of:

Training the first feature extraction module and the second feature extraction module by using the second sample recommendation information and the real recommendation information to obtain a trained first feature extraction module and a trained second feature extraction module;

training a deep learning model by using the first sample recommendation information and the real recommendation information, and obtaining an information recommendation model comprises:

and under the condition that model parameters of the trained first feature extraction module and the trained second feature extraction module are kept unchanged, training the association module by using the first sample recommendation information and the real recommendation information to obtain the information recommendation model.

12. The method of claim 11, wherein the first feature extraction module comprises a converter-based bi-directional encoding model, the second feature extraction module comprises a factorizer model, and the correlation module comprises a deep neural network model.

13. An information recommendation apparatus, comprising:

14. A training device of an information recommendation model, comprising:

the second acquisition module is used for acquiring sample user information corresponding to a sample user, wherein the sample user information comprises a sample historical browsing track information set, a sample user profile information set and real recommendation information;

the third processing module is used for processing the sample historical browsing track information set to obtain a sample historical browsing track characteristic sequence;

15. An electronic device, comprising:

one or more processors;

a memory for storing one or more instructions,

wherein the one or more instructions, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1 to 9 or any of claims 10 to 12.

16. A computer readable storage medium having stored thereon executable instructions which when executed by a processor cause the processor to implement the method of any of claims 1 to 9 or any of claims 10 to 12.

17. A computer program product comprising computer executable instructions for implementing the method of any one of claims 1 to 9 or any one of claims 10 to 12 when executed.