CN112163164B - User tag determining method and related device - Google Patents

User tag determining method and related device Download PDF

Info

Publication number
CN112163164B
CN112163164B CN202011109112.2A CN202011109112A CN112163164B CN 112163164 B CN112163164 B CN 112163164B CN 202011109112 A CN202011109112 A CN 202011109112A CN 112163164 B CN112163164 B CN 112163164B
Authority
CN
China
Prior art keywords
user
content
determining
tag
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011109112.2A
Other languages
Chinese (zh)
Other versions
CN112163164A (en
Inventor
赵猛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tencent Technology Shenzhen Co Ltd
Original Assignee
Tencent Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tencent Technology Shenzhen Co Ltd filed Critical Tencent Technology Shenzhen Co Ltd
Priority to CN202011109112.2A priority Critical patent/CN112163164B/en
Publication of CN112163164A publication Critical patent/CN112163164A/en
Application granted granted Critical
Publication of CN112163164B publication Critical patent/CN112163164B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Strategic Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Economics (AREA)
  • Computational Linguistics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Game Theory and Decision Science (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The embodiment of the application discloses a user tag determining method and a related device, which at least relate to deep learning in an artificial intelligence technology, data parallel processing in a cloud computing technology and the like, acquire user information and content tags corresponding to a target user, then determine tag features corresponding to the content tags, determine user features corresponding to the user information through a user feature sub-model in a user prediction model, then determine whether the target user can establish an association relationship with target content according to the tag features and the user features, and if so, allocate the content tags to the target user. The method for determining the user characteristics by adopting the attention mechanism ensures that the user prediction model can pay more attention to effective information related to the prediction result in the user information and pay less attention to interference information unrelated to the prediction result in the user information, thereby improving the accuracy of determining the user label.

Description

User tag determining method and related device
Technical Field
The present disclosure relates to the field of artificial intelligence, and in particular, to a method and an apparatus for determining a user tag.
Background
With the development of the internet, users can be labeled according to the interaction behavior of the users aiming at different contents, for example, users can be labeled with a game after clicking and watching game advertisements. According to the user tag, the content possibly interested by the user can be recommended for the user in a targeted manner, and therefore the use experience of the user can be effectively improved. Therefore, how to determine a user tag is a matter of investigation.
Disclosure of Invention
In order to solve the technical problems, the application provides a user tag determining method and a related device, and the purpose of determining the user tag is achieved.
In one aspect, an embodiment of the present application provides a method for determining a user tag, where the method includes:
acquiring user information and content labels corresponding to a target user; the content tag is used for identifying one type of target content;
determining tag characteristics corresponding to the content tags, and determining user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model; the user characteristic sub-model comprises n hidden characteristic layers, wherein the input characteristics of the (k+1) th hidden characteristic layer comprise original characteristics corresponding to the user information and output characteristics of the (k) th hidden characteristic layer, and the output characteristics of the (k+1) th hidden characteristic layer are determined according to the importance information of the output characteristics of the (k) th hidden characteristic layer relative to the original characteristics; k+1 is not greater than n;
Determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics;
if yes, the content label is distributed to the target user.
On the other hand, an embodiment of the present application provides a user tag determining apparatus, where the apparatus includes an obtaining unit, a determining unit, and an allocating unit:
the acquisition unit is used for acquiring user information and content labels corresponding to the target users; the content tag is used for identifying one type of target content;
the determining unit is used for determining the label characteristics corresponding to the content labels and determining the user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model; the user characteristic sub-model comprises n hidden characteristic layers, wherein the input characteristics of the (k+1) th hidden characteristic layer comprise original characteristics corresponding to the user information and output characteristics of the (k) th hidden characteristic layer, and the output characteristics of the (k+1) th hidden characteristic layer are determined according to the importance information of the output characteristics of the (k) th hidden characteristic layer relative to the original characteristics; k+1 is not greater than n;
the determining unit is further configured to determine, according to the tag feature and the user feature, whether the target user will establish an association relationship with the target content; if yes, triggering the distribution unit;
The distribution unit is used for distributing the content labels to the target users.
In another aspect, an embodiment of the present application provides an apparatus for determining a user tag, where the apparatus includes a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to perform the method of the above aspect according to instructions in the program code.
In another aspect, embodiments of the present application provide a computer-readable storage medium for storing a computer program for performing the method described in the above aspect.
In another aspect, embodiments of the present application provide a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the method described in the above aspect.
According to the technical scheme, the user information corresponding to the target user and the content tag are acquired, wherein the content tag is used for identifying one type of target content. Then, determining the label characteristics corresponding to the content labels, determining the user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model, and then determining whether the target user can establish an association relationship with the target content according to the label characteristics and the user characteristics, if so, indicating that the target user is interested in the target content identified by the content labels to a larger extent, thereby being capable of distributing the content labels to the target user. Based on the method, the matching problem between the user and the tag is converted into the prediction problem whether the user can establish the association relation with the content, so that the purpose of determining the corresponding tag for the content by the user is achieved according to the prediction result. In the process of determining the user features by using the user feature sub-model, the output features of the (k+1) th implicit feature layer are determined according to the importance information of the kth implicit feature layer relative to the original features corresponding to the user information, and the mode of determining the user features by adopting the attention mechanism ensures that the user prediction model can pay more attention to the effective information related to the prediction result in the user information and pay less attention to the interference information irrelevant to the prediction result in the user information when the user features are used for predicting whether the association relationship between the user and the content exists or not, so that the matching degree of the prediction result and the actual result of the association relationship between the user and the content is higher, and the accuracy of determining the user label is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive faculty for a person skilled in the art.
Fig. 1 is an application scenario schematic diagram of a user tag determining method provided in an embodiment of the present application;
fig. 2 is a flow chart of a method for determining a user tag according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a user prediction model based on a dual-tower structure according to an embodiment of the present application;
FIG. 4 is a schematic flow chart of calculating output features of an implicit feature layer according to an embodiment of the present application;
FIG. 5 is a flowchart illustrating another embodiment of calculating output features of an implicit feature layer;
FIG. 6 is a flowchart illustrating another embodiment of calculating output features of an implicit feature layer;
FIG. 7 is a flowchart of calculating user characteristics according to an embodiment of the present application;
FIG. 8 is a flowchart of a user prediction model training method according to an embodiment of the present application;
fig. 9 is an application scenario schematic diagram of another method for determining a user tag according to an embodiment of the present application;
fig. 10 is a schematic structural diagram of a user tag determining apparatus according to an embodiment of the present application;
fig. 11 is a schematic structural diagram of a server according to an embodiment of the present application;
fig. 12 is a schematic structural diagram of a terminal device according to an embodiment of the present application.
Detailed Description
Embodiments of the present application are described below with reference to the accompanying drawings.
In order to achieve the purpose of determining the user tag, the application provides a user tag determining method and a related device.
The user tag determination method provided by the embodiment of the application is realized based on artificial intelligence, wherein the artificial intelligence (Artificial Intelligence, AI) is a theory, a method, a technology and an application system which simulate, extend and extend human intelligence by using a digital computer or a machine controlled by the digital computer, sense environment, acquire knowledge and acquire an optimal result by using the knowledge. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision.
The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.
In the embodiment of the application, the mainly related artificial intelligence software technology comprises the directions of natural language processing, machine learning/deep learning and the like. For example, text preprocessing (Text preprocessing), semantic understanding (Semantic understanding) in natural language processing (Nature Language processing, NLP) may be involved, and Deep Learning (Deep Learning) in Machine Learning (ML) may be involved, including various types of artificial neural networks (Artificial Neural Network, ANN).
The user tag determining method provided by the application can be applied to user tag determining equipment with data processing capability, such as terminal equipment and a server. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud computing service. The terminal may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal device and the server may be directly or indirectly connected through wired or wireless communication, which is not limited herein.
The user tag determination device may be provided with the capability to perform natural language processing (Nature Language processing, NLP), an important direction in the fields of computer science and artificial intelligence. It is studying various theories and methods that enable effective communication between a person and a computer in natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Thus, the research in this field will involve natural language, i.e. language that people use daily, so it has a close relationship with the research in linguistics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic questions and answers, knowledge graph techniques, and the like. In the embodiment of the application, the user tag determining device can process the user information and the content tag through text preprocessing, semantic understanding and other technologies in natural language processing.
The user tag determination device may be machine learning capable. Machine learning is a multi-domain interdisciplinary, involving multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and the like. It is specially studied how a computer simulates or implements learning behavior of a human to acquire new knowledge or skills, and reorganizes existing knowledge structures to continuously improve own performance. Machine learning is the core of artificial intelligence, a fundamental approach to letting computers have intelligence, which is applied throughout various areas of artificial intelligence. Machine learning and deep learning typically involve techniques such as artificial neural networks.
The artificial intelligence model adopted in the user tag determining method provided by the embodiment of the application mainly relates to application to a neural network, and the corresponding tag of the user is determined through the neural network.
In addition, the user tag determining device provided by the embodiment of the application further has cloud computing capability. Cloud computing (closed computing) refers to the delivery and usage mode of an IT infrastructure, meaning that required resources are obtained in an on-demand, easily scalable manner through a network; generalized cloud computing refers to the delivery and usage patterns of services, meaning that the required services are obtained in an on-demand, easily scalable manner over a network. Such services may be IT, software, internet related, or other services. Cloud Computing is a product of fusion of traditional computer and network technology developments such as Grid Computing (Grid Computing), distributed Computing (Distributed Computing), parallel Computing (Parallel Computing), utility Computing (Utility Computing), network storage (Network Storage Technologies), virtualization (Virtualization), load balancing (Load balancing), and the like.
With the development of the internet, real-time data flow and diversification of connected devices, and the promotion of demands of search services, social networks, mobile commerce, open collaboration and the like, cloud computing is rapidly developed. Unlike the previous parallel distributed computing, the generation of cloud computing will promote the revolutionary transformation of the whole internet mode and enterprise management mode in concept.
In the embodiment of the application, the user tag determining device may process the user information and the content tag by using a cloud computing technology, so as to determine whether the content tag is allocated to the user according to the processed information.
It should be noted that, information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use, and processing of related data are required to comply with related laws and regulations and standards of related countries and regions.
In order to facilitate understanding of the technical solution of the present application, the method for determining a user tag provided in the embodiment of the present application is introduced below by using a server as a user tag determining device in combination with an actual application scenario.
Referring to fig. 1, fig. 1 is an application scenario schematic diagram of a user tag determining method provided in an embodiment of the present application. In the application scenario shown in fig. 1, a server 101 is included as a user tag determination device.
In the application process, the server 101 may acquire the user information and the content tag corresponding to the user a. Wherein the user information is used to identify information related to user a and the content tag is used to identify a type of targeted content. In the scenario shown in fig. 1, the target content is a game-like advertisement, and the corresponding content tag is "game".
Then, the server 101 may perform feature extraction on the content tag to obtain a corresponding tag feature, and determine a user feature corresponding to the user information by using a user feature sub-model in the user prediction model.
The user prediction model is pre-deployed in the server 101, and includes a user feature sub-model, which takes user information as input, performs feature extraction on the user information, and outputs user features.
As shown in fig. 1, the user feature sub-model includes n hidden feature layers, and for the (k+1) th hidden feature layer of the n hidden feature layers, the input features include original features X corresponding to the user information 0 And the output feature X of the kth implicit feature layer k And the output feature X of the (k+1) th implicit feature layer k+1 Output feature X according to the kth implicit feature layer k Relative to original feature X 0 Wherein the importance information identifies the importance information in the original feature X 0 Presence ofIn the case of (a), the output feature X of the kth implicit feature layer k Relative to original feature X 0 Wherein k+1 is not greater than n.
According to the method for determining the output characteristics of each implicit characteristic layer based on the attention mechanism, when the user prediction model predicts the association relation between the user and the target content by using the user characteristics, the user characteristics can be selectively learned, namely, more effective information with higher correlation degree with the prediction result in the user information is focused, less interference information with lower correlation degree with the prediction result in the user information is focused, the influence of the interference information on the prediction result is reduced, the coincidence degree of the prediction result and the actual result of whether the association relation exists between the user and the content is improved, and the accuracy of determining the user label is further improved.
Then, the user prediction model in the server 101 determines whether the user a will establish an association relationship with the target content according to the user characteristics and the tag characteristics. If it is determined that the association relationship between the user A and the target content is established, the user A has a higher interest in the target content, so that the content tag corresponding to the target content can be assigned to the user A as the tag of the user A.
The user tag determining method converts the matching problem between the user and the tag into the prediction problem of whether the user can establish the association relation with the content, so that the purpose of determining the corresponding tag for the content by the user is achieved according to the prediction result. In the process of extracting the user characteristics by using the user sub-model in the user prediction model, an attention mechanism is adopted, so that the user prediction model selectively learns the user characteristics, the degree of coincidence between a prediction result and an actual result of whether an association relation exists between a user and content is improved, and the accuracy of determining the user label is further improved.
Referring to fig. 2, fig. 2 is a flowchart of a method for determining a user tag according to an embodiment of the present application. As shown in fig. 2, the user tag determination method includes the steps of:
S201: and acquiring user information and content labels corresponding to the target users.
In practical application, the user information and the content label corresponding to the target user can be obtained from the content log, and the user information and the content label of the target user can be collected in real time. Wherein the user information is used to identify information related to the target user, including but not limited to: basic attributes of the target user (such as age, gender and the like), historical behavior data of the target user for the content (such as behavior data generated by historical clicking of the user for viewing the content), statistical characteristics of the target user for the content (such as historical labels of the user for the content) and the like.
The content tag is used to identify a type of target content. Wherein, the target content refers to content which can be watched or listened to by a user, and the existence form of the target content can include one or more of combination of text, image, audio and video, such as: advertising, news, etc. For a certain type of content, the content tag may be used for identification. Such as: for a game advertisement, its corresponding content tag is "game".
S202: and determining the label characteristics corresponding to the content labels, and determining the user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model.
In practical application, feature extraction can be performed on the content tag by using a feature extraction model in the artificial intelligence, so as to obtain the tag feature corresponding to the content tag. The feature extraction model may be a neural network model with any result, and is not limited in this regard. For the user information, the user characteristic sub-model in the user prediction model can be utilized to extract the characteristics of the user information, and the content characteristics corresponding to the user information are obtained.
In the embodiment of the application, the user information and the content labels can be extracted by adopting a user prediction model with a double-tower structure. As shown in fig. 3, the user prediction model includes a tag feature sub-model for performing feature extraction on the content tag and the above-mentioned user feature sub-model for performing feature extraction on the user information.
For the tag feature sub-model, a convolutional neural network (Convolutional Network Nerual, CNN) model structure may be used, or other network model structure may be used, which is not limited in this regard. The tag features are output with the content tags as inputs to the model.
For the user feature sub-model, a deep neural network (Deep Network Nerual, DNN) model structure may be employed, considering more user information. User information is taken as input. Then, the user information is subjected to vector representation (emmbedding) to obtain the original characteristic X corresponding to the user information 0 If the original characteristic X 0 Includes m feature numbers, and each feature has dimension D, then the original feature X 0 Can be expressed as [ m, D]。
As shown in fig. 3, the user feature sub-model includes n implicit feature layers. For the (k+1) th implicit feature layer, its input features include original feature X 0 And the output feature X of the kth implicit feature layer k And the output feature X of the (k+1) th implicit feature layer k+1 Output feature X according to the kth implicit feature layer k Relative to original feature X 0 Is determined by the importance information of (a), see in particular fig. 1.
Wherein the importance information identifies the importance information in the original feature X 0 On an existing basis, the output characteristic X of the kth implicit characteristic layer k Relative to original feature X 0 Is of importance. In the feature extraction process of n implicit feature layers, attention (attention) mechanisms are introduced, so that when the user prediction model predicts the user features, the user prediction model can selectively learn the user features, namely, more attention is paid to effective information with higher correlation degree with the prediction results, less attention is paid to interference information with lower correlation degree with the prediction results, the influence of the interference information on the prediction results is reduced, the coincidence degree of the prediction results with actual results for whether association relation exists between the user and the content is improved, and the accuracy of determining the user labels is further improved.
As shown in FIG. 3, considering that a plurality of user information is independently input into the user feature sub-model, such as the age in FIG. 3The user information is independent of each other, and when the user information is extracted by features, the combination of the features is difficult to achieve. In view of this, in the present embodiment, the original feature X is utilized 0 And the output feature X of the kth implicit feature layer k Determining the output feature X of the (k+1) th implicit feature layer k+1 When the original characteristic X is 0 And the output feature X of the kth implicit feature layer k Performing feature crossing, and taking the feature obtained after the feature crossing as the output feature X of the (k+1) th hidden feature layer k+1 。
Specifically, if the original characteristic X 0 Comprises m elements, denoted as [ e ] 1 ,e 2 ,…,e q ,…,e m ]First element e q Is the original characteristic X 0 The output feature X of the kth hidden feature layer of any one of the m elements included k Includes H k The individual elements are denoted as
Referring to fig. 4, first according to the first element e q And H k Determining H in importance information k The individual element is relative to the first element e q Sub-importance information alpha i,q Wherein i takes the values of 1,2, … and H k Abbreviated as alpha :,q . By H k H in the individual elements k Individual elementsFor example, H k Individual element->Relative to the first element e q Sub-importance information->The following formula can be used for calculation:
wherein the sub-importance informationIs based on e q In case of query +.>For the attention score (attention score) at key value (key), this calculation process can be regarded as the feature element e q And->Crossing between, whereas the attention score identifies the characteristic element +.>Relative characteristic element e q May also be referred to as importance weight.
Based on the above formula, the sub-importance information α of the first element can be determined i,q As shown in fig. 4, α 1,q ＝0.1,α 2,q ＝0.2,,,,,Then, the sub-importance information alpha of the first element is determined i,q And H is k Cross information E of individual elements i,q Wherein i takes the values of 1,2, … and H k Abbreviated as E :,q . For E :,q Can be denoted as Z k+1,q By H k H in the individual elements k Individual element->For example, the sub-importance information of the first element +.>And H th k Cross information of individual elementsThe following formula can be used for calculation:
then, according to the original feature X 0 The m elements [ e ] included 1 ,e 2 ,…,e q ,…,e m ]Respectively corresponding cross information E i,j Determination of the Cross-over feature Z of the (k+1) th implicit feature layer k+1,j Wherein i takes the values of 1,2, … and H k J has the value of 1,2, …, m, abbreviated as Z k+1,: . For Z k+1,: Can be denoted as Z k+1 Dimension is H k * m.d, fig. 5 shows the original feature X 0 [m,D]Output feature X with kth implicit feature layer k [H k ,D]Feature intersection based on attention mechanism to obtain intersection feature Z of kth+1th implicit feature layer k+1 [H k ,m,D]Is a schematic diagram of (a).
Since the cross feature has higher dimension, it can cause dimension explosion, so the cross feature Z can be used k+1 [H k ,m,D]And carrying out weighted aggregation. As shown in fig. 6, for cross feature Z k+1 [H k ,m,D]After weighted aggregation, the output characteristic X after the dimension reduction of the (k+1) th implicit characteristic layer is obtained k+1 [H k+1 ,D]The formula is as follows:
X k+1 ＝w ij E i,j ,0≤i≤H k ,0≤j≤m
wherein w is ij Representing crossing information E i,j Weighting coefficient w of (2) ij Can be preset.
From the above, the output features X of the n hidden feature layers can be determined respectively 1 ,X 2 ,…,X n Wherein the 1 st implicit feature layer outputs the feature X 1 The original feature X can be utilized 0 Crossing itself, the specific process is shown in fig. 7.
Based on the above, on the basis of explicit feature intersection (i.e. the feature of intersection is known), an attention mechanism is introduced, so that the model can selectively learn the user features, the influence of interference information on a prediction result is reduced, the prediction accuracy of whether an association relationship exists between a user and content is improved, and the accuracy of determining the user tag is improved.
Based on the above, the output features X of the n hidden feature layers can be obtained 1 ,X 2 ,…,X n And determining the user characteristics corresponding to the user information. Specifically, the output features X of n hidden feature layers are first outputted 1 ,X 2 ,…,X n And respectively performing connection operation (concat) to obtain n one-dimensional features, and then performing full connection operation on the n one-dimensional features to obtain the user features corresponding to the user information.
In the process of determining the user characteristics, the output characteristics of each implicit characteristic layer are combined, so that the characteristics determined by each layer according to the importance information are reserved in the user characteristics, and the user prediction model can predict whether the association relationship exists between the user and the content according to the global information.
S203: and determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics. If yes, S204 is executed.
After determining the tag feature corresponding to the content tag and the user feature corresponding to the user information based on the S202, it may be determined whether the target user will establish an association relationship with the target content identified by the content tag according to the two features. If the target user is determined to establish an association relationship with the target content, the target user is shown to be interested in the target content to a larger degree; if the target user is determined not to establish the association relation with the target content, the target user is indicated to be interested in the target content to a smaller degree.
In one possible implementation, if the user prediction model further includes a predictor model, the user feature and the tag feature may be used as inputs of the predictor model to output a prediction result of whether the target user will establish an association with the target content, as shown in fig. 3. The model structure of the predictor model may be set according to the actual application scenario, and is not limited herein.
The user prediction model provided by the embodiment, or also called as a attention mechanism-based feature interaction (Deep Attention Network, deep attint) model, realizes parallel processing of user information and content labels through the user feature sub-model and the content feature sub-model of the double-tower structure, improves the processing efficiency of feature extraction, predicts whether a user can establish an association relationship with the content by combining with the prediction sub-model, and achieves the purpose of determining the user labels according to the prediction result.
For the process of determining the association relationship, a possible implementation manner is provided, namely, similarity information between the user characteristics and the label characteristics is calculated. And then judging whether the similarity information meets the association condition or not, and determining whether the target user can establish an association relationship with the target content or not. If the similarity information meets the association condition, determining that the target user can establish an association relationship with the target content; if the similarity information does not meet the association condition, determining that the target user cannot establish an association relationship with the target content.
In practical applications, cosine similarity (cosine similarity) may be used to calculate similarity information between the user features and the tag features, the similarity information identifying the degree of similarity between the user features and the tag features. The higher the similarity between the user features and the tag features is, the greater the possibility that the target user has an association relationship with the target content is indicated; the lower the degree of similarity between the user features and the tag features, the less likely the target user has an association with the target content.
The association condition identifies a condition that the similarity information between the user feature and the tag feature needs to satisfy when the target user has an association relationship with the target content. If the similarity information meets the association condition, the similarity information between the user characteristics and the tag characteristics is indicated to reach the condition required by the association relation between the target user and the target content, so that the association relation between the target user and the target content can be determined. If the similarity information does not meet the association condition, the similarity information between the user characteristics and the tag characteristics does not meet the condition required by the association relation between the target user and the target content, so that the target user can be determined not to establish the association relation with the target content. In practical application, the association condition may be preset, or may be adaptively adjusted according to the practical application scenario, which is not limited in any way.
S204: and distributing the content label to the target user.
If it is determined that the target user can establish an association relationship with the target content, it indicates that the target user is interested in the target content to a greater extent, and in this case, the content tag may be allocated to the target user. If it is determined that the target user does not establish an association relationship with the target content, it indicates that the target user is interested in the target content to a small extent, and in this case, the content tag is not allocated to the target user.
In practical application, a plurality of content tags may be obtained, and after similarity information between tag features corresponding to the plurality of content tags and user features is determined based on the steps, content tags (that is, topK) with similarity of the first K may be selected as tags of the target user.
It can be understood that the user prediction model provided in the above embodiment can predict whether the association relationship between the user and the content is established, so that in practical application, the steps of acquiring the interaction behavior data between the user and the content and carrying out statistics induction according to the interaction behavior data in the process of determining the user tag based on a statistical manner are avoided, the operation flow of determining the user tag is simplified, and the data processing capacity is reduced.
In view of this, in determining the user tag by using the user prediction model, user information and content tags corresponding to the target user in the first period of time may be obtained, and according to user features corresponding to the user information and tag features corresponding to the content tags, whether the target user will establish an association relationship with the target content in the second period of time may be predicted. Wherein the first period of time is earlier than the second period of time.
The historical data is used for predicting whether the association relationship between the user and the content is established later, so that the problem of characteristic leakage caused by predicting by using the data in a time period is avoided, and the accuracy of predicting the association relationship between the user and the content is ensured.
The user tag determining method provided in the above embodiment obtains user information corresponding to a target user and a content tag, where the content tag is used to identify a type of target content. Then, determining the label characteristics corresponding to the content labels, determining the user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model, and then determining whether the target user can establish an association relationship with the target content according to the label characteristics and the user characteristics, if so, indicating that the target user is interested in the target content identified by the content labels to a larger extent, thereby being capable of distributing the content labels to the target user. Based on the method, the matching problem between the user and the tag is converted into the prediction problem whether the user can establish the association relation with the content, so that the purpose of determining the corresponding tag for the content by the user is achieved according to the prediction result. In the process of determining the user features by using the user feature sub-model, the output features of the (k+1) th implicit feature layer are determined according to the importance information of the kth implicit feature layer relative to the original features corresponding to the user information, and the mode of determining the user features by adopting the attention mechanism ensures that the user prediction model can pay more attention to the effective information related to the prediction result in the user information and pay less attention to the interference information irrelevant to the prediction result in the user information when the user features are used for predicting whether the association relationship between the user and the content exists or not, so that the matching degree of the prediction result and the actual result of the association relationship between the user and the content is higher, and the accuracy of determining the user label is improved.
Aiming at the user prediction model provided by the embodiment, the embodiment of the application provides a training method of the user prediction model.
Referring to fig. 8, fig. 8 is a flowchart of a user prediction model training method according to an embodiment of the present application. As shown in fig. 8, the training method includes:
s801: and determining a training sample according to sample user information and sample content labels corresponding to the sample users.
In practical application, the training sample can be determined according to sample user information and sample content labels corresponding to the sample user. Wherein the sample user information identifies information related to the sample user, including, but not limited to: basic attributes of the sample user, historical behavior data of the sample user for the sample content, and statistical tags of the sample user for the sample content. The sample content tag identifies one type of sample content. The training sample comprises sample user information, sample content labels and identification of whether the sample user establishes an association relationship with sample content. In practical application, the existence of interaction behavior between the sample user and the sample content can be regarded as that the association relationship between the sample user and the sample content is established.
S802: and training the user prediction model according to the training sample.
In the training process, determining sample user characteristics corresponding to sample users through a user characteristic sub-model, determining sample tag characteristics corresponding to sample content tags through a tag characteristic sub-model, and correspondingly adjusting model parameters of a user prediction model if a result of whether a user determined by a prediction sub-model can establish an association relationship with sample content is inconsistent with the identification according to the sample user characteristics and the sample tag characteristics.
By the user prediction model training method, the user prediction model can be trained, so that the user prediction model can learn the characteristics in combination with the attention mechanism mode on the basis of characteristic intersection, the user prediction model is enabled to have the capability of predicting whether the association relation between the user and the content exists, and therefore the user label can be determined by using the user prediction model.
In order to verify the effect of the user tag determination method provided in the embodiment of the present application, a comparison experiment is performed on several models commonly used in the related art and the attention mechanism-based feature interaction (deep int) model provided in the above embodiment, respectively, on the internal data sets of the public data set and the user prediction model. The models involved in the comparison experiment include a generalized depth (wire & Deep, LR) model, a Deep-decomposition (FM) model, a Deep-Cross (DCN) model, an extreme Deep factorization (CIA) model, and a Deep neural Network (Deep Network Nerual, DNN) model.
Comparative experiment one:
the data set employed: criterion discloses a dataset.
Data set profile: about 40M samples, 23 feature sets, discrete features comprising 1.5M.
Data set partitioning: randomly dividing into 80% training set, 10% verification set and 10% test set.
Data preprocessing: for continuous features, to reduce their discreteness, y=log is done 2 x, and all features with global frequency less than 5 times are set as absense.
The results are shown in Table 1:
comparison experiment II:
data set: social software user interest category label modeling dataset
Data set profile: the training data of 3 hundred million, 2000w test sample, 20 characteristic groups, label strategy is multi-label hit expansion, and when one exposure hits N hit orientation fields, the training samples are split into N training samples.
The results are shown in Table 2:
as can be seen by combining the above three parameters of Accuracy (Auc), loss (loglos), and the ratio of improvement compared with the LR model in table 1 and table 2, the attention mechanism-based feature interaction (deep int) model provided in the embodiment of the present application achieves a better effect than the model in the related art.
It should be noted that, the method for determining the user tag provided in the above embodiment may be applied to various different scenarios, for example, advertisement recall through the user tag, video recommendation, news recommendation, etc. to the user through the user tag.
For better understanding, a method for determining a user tag provided in the embodiments of the present application is described below in conjunction with an application scenario of advertisement recall. Generally, the advertisement system mainly comprises main flows of directional recall, coarse ranking, fine ranking and the like, wherein the directional recall is mainly realized by marking a specific label on a user, and then recalling a specific crowd by means of the label. The coarse ranking is to roughen the oriented crowd, and deliver the refined ranking module to carefully select and display. In the application, the user is labeled by using the user label determining method provided by the embodiment, so that the directional recall of the advertisement is realized.
Referring to fig. 9, fig. 9 is an application scenario schematic diagram of a user tag determining method provided in an embodiment of the present application. In the application scenario shown in fig. 9, three processes of sample extraction, model training and user label assignment are included. In the process of determining the user tag, the target content is the advertisement, whether the association relationship exists between the user and the advertisement can be converted into the interactive behavior of whether the user clicks to view the advertisement,
in the sample extraction process, user information (901) and advertisement logs (902) may be collected first. The user information comprises user basic attributes, historical behavior data of clicking advertisements by the user and statistical data of the advertisements by the user. Advertisement data (user, tag, ad, exposure/click, time) includes user identification (user), advertisement Tag (Tag), whether advertisement (Ad) advertisement exposure is clicked by a user, and advertisement data storage time.
Then, a training sample (903) is constructed based on the advertisement data included in the advertisement log, wherein the training sample (user, tag, label) includes a user identifier (user), an advertisement Tag (Tag), and an identifier (label) for identifying whether the advertisement exposure is clicked by the user. Wherein the advertisement exposure is clicked by the user as a positive sample, i.e. the sign is marked as 1, otherwise as a negative sample, the sign is marked as 0. Based thereon, a plurality of training samples (904) are acquired, such as: (user 1, tag1, label 1), (user 2, tag2, label 2), (user 3, tag3, label 3).
In the model training process, the user predictive model is first trained using user information, advertisement labels and identifications (i.e., the training samples described above) (906). Specifically, the user information, the advertisement tag and the identification are taken as input, and the result of whether the user clicks the advertisement, namely, P (click|user, tag) is output. And if the prediction result is inconsistent with the identification, performing parameter adjustment on the user prediction model. In the process of predicting whether the user clicks the advertisement by using the user information and the training sample, the storage time of the advertisement data is required to be combined, so that the acquisition time of the sample data, namely the characteristic time t1, is ensured before the prediction time t2 of whether the user clicks the advertisement, and thus the characteristic leakage is avoided (904).
In the distribution process, whether the user clicks the advertisement or not is predicted by utilizing the trained user prediction model, and a label is distributed to the user according to a prediction result (907), so that the directional recall (908) of the advertisement is realized through the label corresponding to the user;
according to the method for determining the user tag, the corresponding advertisement tag is distributed to the user, so that the efficiency of distributing the tag to the user is improved, based on the fact, the distributed user tag is used for realizing the directional recall of the advertisement, and the advertisement recall efficiency is improved.
The embodiment of the application also provides a user tag determining device aiming at the user tag determining method provided by the embodiment.
Referring to fig. 10, fig. 10 is a schematic diagram of a user tag determining apparatus according to an embodiment of the present application. As shown in fig. 10, the user tag determination apparatus 1000 includes an acquisition unit 1001, a determination unit 1002, and an allocation unit 1003:
the acquiring unit 1001 is configured to acquire user information and a content tag corresponding to a target user; the content tag is used for identifying one type of target content;
the determining unit 1002 is configured to determine a tag feature corresponding to the content tag, and determine a user feature corresponding to the user information through a user feature sub-model in a user prediction model; the user characteristic sub-model comprises n hidden characteristic layers, wherein the input characteristics of the (k+1) th hidden characteristic layer comprise original characteristics corresponding to the user information and output characteristics of the (k) th hidden characteristic layer, and the output characteristics of the (k+1) th hidden characteristic layer are determined according to the importance information of the output characteristics of the (k) th hidden characteristic layer relative to the original characteristics; k+1 is not greater than n;
The determining unit 1002 is further configured to determine, according to the tag feature and the user feature, whether the target user will establish an association relationship with the target content; if yes, triggering the distribution unit;
the allocation unit 1003 is configured to allocate the content tag to the target user.
In one possible implementation, if the first element is any element in the original features, the output features of the kth implicit feature layer include H k A number of elements, the determining unit 1002 is configured to:
according to the first element and the H k A plurality of elements for determining the H in the importance information k Sub-importance information of the individual elements relative to the first element;
determining sub-importance information of the first element and the H k Intersection information of the individual elements;
and determining the output characteristics of the (k+1) th implicit characteristic layer according to the intersection information respectively corresponding to the elements included in the original characteristics.
In a possible implementation manner, the determining unit 1002 is configured to:
determining n output features corresponding to the n hidden feature layers through the user feature sub-model;
and determining the user characteristics corresponding to the user information according to the n output characteristics.
In a possible implementation manner, the determining unit 1002 is configured to:
calculating similarity information between the user features and the tag features;
if the similarity information meets the association condition, determining that the target user can establish an association relationship with the target content;
and if the similarity information does not meet the association condition, determining that the target user cannot establish an association relationship with the target content.
In a possible implementation manner, the determining unit 1002 is configured to:
determining the label characteristics corresponding to the content labels through a label characteristic sub-model in the user prediction model;
and determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics through a predictor model in the user prediction model.
In a possible implementation manner, the determining unit 1002 is further configured to:
determining a training sample according to sample user information and sample content labels corresponding to sample users; the sample content tag is used for identifying one type of sample content; the training sample comprises the sample user information, a sample content label and an identifier of whether the sample user can establish an association relationship with the sample content;
The device further comprises a training unit:
the training unit is used for training the user prediction model according to the training sample;
in the training process, determining sample user characteristics corresponding to the sample user through the user characteristic sub-model, determining sample tag characteristics corresponding to the sample content tag through the tag characteristic sub-model, and correspondingly adjusting model parameters of the user prediction model if a result of whether the user determined through the prediction sub-model can establish an association relationship with the sample content is inconsistent with the identifier according to the sample user characteristics and the sample tag characteristics.
In a possible implementation manner, the obtaining unit 1001 is configured to obtain the user information and the content tag in a first period of time;
the determining whether the target user can establish an association relationship with the target content according to the user characteristics and the tag characteristics comprises:
determining whether the target user can establish an association relationship with target content in a second time period according to the user characteristics and the tag characteristics; the first period of time is earlier than the second period of time.
The user tag determining device provided in the above embodiment obtains user information corresponding to a target user and a content tag, where the content tag is used to identify a type of target content. Then, determining the label characteristics corresponding to the content labels, determining the user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model, and then determining whether the target user can establish an association relationship with the target content according to the label characteristics and the user characteristics, if so, indicating that the target user is interested in the target content identified by the content labels to a larger extent, thereby being capable of distributing the content labels to the target user. Based on the method, the matching problem between the user and the tag is converted into the prediction problem whether the user can establish the association relation with the content, so that the purpose of determining the corresponding tag for the content by the user is achieved according to the prediction result. In the process of determining the user features by using the user feature sub-model, the output features of the (k+1) th implicit feature layer are determined according to the importance information of the kth implicit feature layer relative to the original features corresponding to the user information, and the mode of determining the user features by adopting the attention mechanism ensures that the user prediction model can pay more attention to the effective information related to the prediction result in the user information and pay less attention to the interference information irrelevant to the prediction result in the user information when the user features are used for predicting whether the association relationship between the user and the content exists or not, so that the matching degree of the prediction result and the actual result of the association relationship between the user and the content is higher, and the accuracy of determining the user label is improved.
The embodiment of the application also provides equipment for determining the user tag, and the equipment for determining the user tag provided by the embodiment of the application is introduced from the aspect of hardware materialization.
Referring to fig. 11, fig. 11 is a schematic diagram of a server structure provided in an embodiment of the present application, where the server 1400 may vary considerably in configuration or performance, and may include one or more central processing units (central processing units, CPU) 1422 (e.g., one or more processors) and memory 1432, one or more storage media 1430 (e.g., one or more mass storage devices) that store applications 1442 or data 1444. Wherein the memory 1432 and storage medium 1430 can be transitory or persistent storage. The program stored in the storage medium 1430 may include one or more modules (not shown), each of which may include a series of instruction operations on a server. Further, the central processor 1422 may be provided in communication with a storage medium 1430 to perform a series of instruction operations in the storage medium 1430 on the server 1400.
The server 1400 may also include one or more power supplies 1426, one or more wired or wireless network interfaces 1450, one or more input/output interfaces 1458, and/or one or more operating systems 1441, such as Windows Server, mac OS XTM, unixTM, linuxTM, freeBSDTM, and the like.
The steps performed by the server in the above embodiments may be based on the server structure shown in fig. 11.
Wherein, the CPU 1422 is configured to perform the following steps:
acquiring user information and content labels corresponding to a target user; the content tag is used for identifying one type of target content;
determining tag characteristics corresponding to the content tags, and determining user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model; the user characteristic sub-model comprises n hidden characteristic layers, wherein the input characteristics of the (k+1) th hidden characteristic layer comprise original characteristics corresponding to the user information and output characteristics of the (k) th hidden characteristic layer, and the output characteristics of the (k+1) th hidden characteristic layer are determined according to the importance information of the output characteristics of the (k) th hidden characteristic layer relative to the original characteristics; k+1 is not greater than n;
determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics;
if yes, the content label is distributed to the target user.
Optionally, the CPU 1422 may further perform method steps of any specific implementation of the user tag determination method in the embodiments of the present application.
For the user tag determining method described above, the embodiment of the present application also provides a terminal device for determining a user tag, so that the user tag determining method described above is actually implemented and applied.
Referring to fig. 12, fig. 12 is a schematic structural diagram of a terminal device according to an embodiment of the present application. For convenience of explanation, only those portions relevant to the embodiments of the present application are shown, and specific technical details are not disclosed, refer to the method portions of the embodiments of the present application. The terminal device may be any terminal device including a mobile phone, a tablet computer, a personal digital assistant (Personal Digital Assistant, PDA for short), etc., taking the terminal device as an example of the mobile phone:
fig. 12 is a block diagram showing a part of the structure of a mobile phone related to a terminal device provided in an embodiment of the present application. Referring to fig. 12, the mobile phone includes: radio Frequency (RF) circuitry 1510, memory 1520, input unit 1530, display unit 1540, sensor 1550, audio circuitry 1560, wireless fidelity (wireless fidelity, wiFi) module 1570, processor 1580, and power supply 1590. Those skilled in the art will appreciate that the handset configuration shown in fig. 12 is not limiting of the handset and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components.
The following describes the components of the mobile phone in detail with reference to fig. 12:
the RF circuit 1510 may be used for receiving and transmitting signals during a message or a call, and particularly, after receiving downlink information of a base station, the signal is processed by the processor 1580; in addition, the data of the design uplink is sent to the base station. Generally, RF circuitry 1510 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, LNA for short), a duplexer, and the like. In addition, the RF circuitry 1510 may also communicate with networks and other devices through wireless communication. The wireless communication may use any communication standard or protocol, including but not limited to global system for mobile communications (Global System of Mobile communication, GSM for short), general packet radio service (General Packet Radio Service, GPRS for short), code division multiple access (Code Division Multiple Access, CDMA for short), wideband code division multiple access (Wideband Code Division Multiple Access, WCDMA for short), long term evolution (Long Term Evolution, LTE for short), email, short message service (Short Messaging Service, SMS for short), and the like.
The memory 1520 may be used to store software programs and modules, and the processor 1580 implements various functional applications and data processing of the handset by running the software programs and modules stored in the memory 1520. The memory 1520 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, phonebook, etc.) created according to the use of the handset, etc. In addition, memory 1520 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device.
The input unit 1530 may be used to receive input numerical or character information and generate key signal inputs related to user settings and function control of the handset. In particular, the input unit 1530 may include a touch panel 1531 and other input devices 1532. The touch panel 1531, also referred to as a touch screen, may collect touch operations thereon or thereabout by a user (e.g., operations of the user on the touch panel 1531 or thereabout by using any suitable object or accessory such as a finger, a stylus, etc.), and drive the corresponding connection device according to a predetermined program. Alternatively, the touch panel 1531 may include two parts, a touch detection device and a touch controller. The touch detection device detects the touch azimuth of a user, detects a signal brought by touch operation and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends the touch point coordinates to the processor 1580, and can receive and execute commands sent from the processor 1580. In addition, the touch panel 1531 may be implemented in various types such as resistive, capacitive, infrared, and surface acoustic wave. The input unit 1530 may include other input devices 1532 in addition to the touch panel 1531. In particular, other input devices 1532 may include, but are not limited to, one or more of a physical keyboard, function keys (e.g., volume control keys, switch keys, etc.), a trackball, mouse, joystick, etc.
The display unit 1540 may be used to display information input by a user or information provided to the user and various menus of the mobile phone. The display unit 1540 may include a display panel 1541, and optionally, the display panel 1541 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an Organic Light-Emitting Diode (OLED), or the like. Further, the touch panel 1531 may cover the display panel 1541, and when the touch panel 1531 detects a touch operation thereon or thereabout, the touch operation is transferred to the processor 1580 to determine the type of touch event, and then the processor 1580 provides a corresponding visual output on the display panel 1541 according to the type of touch event. Although in fig. 12, the touch panel 1531 and the display panel 1541 are two separate components for implementing the input and input functions of the mobile phone, in some embodiments, the touch panel 1531 may be integrated with the display panel 1541 to implement the input and output functions of the mobile phone.
The handset may also include at least one sensor 1550, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor that may adjust the brightness of the display panel 1541 according to the brightness of ambient light, and a proximity sensor that may turn off the display panel 1541 and/or the backlight when the phone is moved to the ear. As one of the motion sensors, the accelerometer sensor can detect the acceleration in all directions (generally three axes), and can detect the gravity and direction when stationary, and can be used for applications of recognizing the gesture of a mobile phone (such as horizontal and vertical screen switching, related games, magnetometer gesture calibration), vibration recognition related functions (such as pedometer and knocking), and the like; other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc. that may also be configured with the handset are not described in detail herein.
Audio circuitry 1560, a speaker 1561, and a microphone 1562 may provide an audio interface between a user and a cell phone. The audio circuit 1560 may transmit the received electrical signal converted from audio data to the speaker 1561, and be converted into a sound signal by the speaker 1561 for output; on the other hand, the microphone 1562 converts the collected sound signals into electrical signals, which are received by the audio circuit 1560 for conversion into audio data, which is processed by the audio data output processor 1580 for transmission to, for example, another cellular phone via the RF circuit 1510 or for output to the memory 1520 for further processing.
WiFi belongs to a short-distance wireless transmission technology, and a mobile phone can help a user to send and receive emails, browse webpages, access streaming media and the like through a WiFi module 1570, so that wireless broadband Internet access is provided for the user. Although fig. 12 shows WiFi module 1570, it is understood that it is not a necessary component of a cell phone and may be omitted entirely as desired within the scope of not changing the essence of the invention.
The processor 1580 is a control center of the mobile phone, connects various parts of the entire mobile phone using various interfaces and lines, and performs various functions of the mobile phone and processes data by running or executing software programs and/or modules stored in the memory 1520 and invoking data stored in the memory 1520. In the alternative, processor 1580 may include one or more processing units; preferably, the processor 1580 can integrate an application processor and a modem processor, wherein the application processor primarily processes operating systems, user interfaces, application programs, and the like, and the modem processor primarily processes wireless communications. It is to be appreciated that the modem processor described above may not be integrated into the processor 1580.
The handset further includes a power supply 1590 (e.g., a battery) for powering the various components, which may preferably be logically connected to the processor 1580 via a power management system so as to provide for the management of charging, discharging, and power consumption by the power management system.
Although not shown, the mobile phone may further include a camera, a bluetooth module, etc., which will not be described herein.
In an embodiment of the present application, the memory 1520 included in the mobile phone may store program codes and transmit the program codes to the processor.
The processor 1580 included in the mobile phone may execute the user tag determining method provided in the foregoing embodiment according to the instructions in the program code.
The embodiment of the application also provides a computer readable storage medium for storing a computer program for executing the user tag determining method provided in the above embodiment.
Embodiments of the present application also provide a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the user tag determination method provided in various alternative implementations of the above aspects.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, where the above program may be stored in a computer readable storage medium, and when the program is executed, the program performs steps including the above method embodiments; and the aforementioned storage medium may be at least one of the following media: read-only memory (ROM), RAM, magnetic disk or optical disk, etc., which can store program codes.
It should be noted that, in the present specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment is mainly described in a different point from other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, with reference to the description of the method embodiments in part. The apparatus and system embodiments described above are merely illustrative, in which elements illustrated as separate elements may or may not be physically separate, and elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
The foregoing is merely one specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (16)

1. A method of user tag determination, the method comprising:
acquiring user information and content labels corresponding to a target user; the content tag is used for identifying one type of target content;
determining tag characteristics corresponding to the content tags, and determining user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model; the user characteristic sub-model comprises n hidden characteristic layers, wherein the input characteristics of the (k+1) th hidden characteristic layer comprise original characteristics corresponding to the user information and output characteristics of the (k) th hidden characteristic layer, and the output characteristics of the (k+1) th hidden characteristic layer are determined according to the importance information of the output characteristics of the (k) th hidden characteristic layer relative to the original characteristics; k+1 is not greater than n, wherein in the process of determining the user characteristics, the output characteristics of each implicit characteristic layer are combined, so that the characteristics determined by each layer according to the importance information are reserved in the user characteristics, and whether the association relationship between the user and the content exists or not is predicted according to the global information;
Determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics, wherein whether the association relationship exists between the target user and the target content is whether the target user clicks to view the interaction behavior of the target content;
if yes, the content label is distributed to the target user.
2. The method of claim 1, wherein if the first element is any one of the original features, the output features of the kth implicit feature layer include H k The output characteristics of the (k+1) th implicit characteristic layer are determined by:
according to the first element and the H k A plurality of elements for determining the H in the importance information k Sub-importance information of the individual elements relative to the first element;
determining sub-importance information of the first element and the H k Intersection information of the individual elements;
and determining the output characteristics of the (k+1) th implicit characteristic layer according to the intersection information respectively corresponding to the elements included in the original characteristics.
3. The method of claim 1, wherein determining, by the user feature sub-model in the user prediction model, the user feature corresponding to the user information comprises:
Determining n output features corresponding to the n hidden feature layers through the user feature sub-model;
and determining the user characteristics corresponding to the user information according to the n output characteristics.
4. The method of claim 1, wherein determining whether the target user will establish an association with the target content based on the tag characteristic and the user characteristic comprises:
calculating similarity information between the user features and the tag features;
if the similarity information meets the association condition, determining that the target user can establish an association relationship with the target content;
and if the similarity information does not meet the association condition, determining that the target user cannot establish an association relationship with the target content.
5. The method of any one of claims 1-4, wherein determining the tag characteristic corresponding to the content tag comprises:
determining the label characteristics corresponding to the content labels through a label characteristic sub-model in the user prediction model;
the determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics comprises:
And determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics through a predictor model in the user prediction model.
6. The method of claim 5, wherein the method further comprises:
determining a training sample according to sample user information and sample content labels corresponding to sample users; the sample content tag is used for identifying one type of sample content; the training sample comprises the sample user information, a sample content label and an identifier of whether the sample user can establish an association relationship with the sample content;
training the user prediction model according to the training sample;
in the training process, determining sample user characteristics corresponding to the sample user through the user characteristic sub-model, determining sample tag characteristics corresponding to the sample content tag through the tag characteristic sub-model, and correspondingly adjusting model parameters of the user prediction model if a result of whether the user determined through the prediction sub-model can establish an association relationship with the sample content is inconsistent with the identifier according to the sample user characteristics and the sample tag characteristics.
7. The method according to any one of claims 1-4, wherein the obtaining the user information and the content tag corresponding to the target user includes:
acquiring the user information and the content tag in a first time period;
the determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics comprises:
determining whether the target user can establish an association relationship with target content in a second time period according to the user characteristics and the tag characteristics; the first period of time is earlier than the second period of time.
8. A user tag determination apparatus, characterized in that the apparatus comprises an acquisition unit, a determination unit and an allocation unit:
the acquisition unit is used for acquiring user information and content labels corresponding to the target users; the content tag is used for identifying one type of target content;
the determining unit is used for determining the label characteristics corresponding to the content labels and determining the user characteristics corresponding to the user information through a user characteristic sub-model in a user prediction model; the user characteristic sub-model comprises n hidden characteristic layers, wherein the input characteristics of the (k+1) th hidden characteristic layer comprise original characteristics corresponding to the user information and output characteristics of the (k) th hidden characteristic layer, and the output characteristics of the (k+1) th hidden characteristic layer are determined according to the importance information of the output characteristics of the (k) th hidden characteristic layer relative to the original characteristics; k+1 is not greater than n, wherein in the process of determining the user characteristics, the output characteristics of each implicit characteristic layer are combined, so that the characteristics determined by each layer according to the importance information are reserved in the user characteristics, and whether the association relationship between the user and the content exists or not is predicted according to the global information;
The determining unit is further configured to determine, according to the tag feature and the user feature, whether the target user will establish an association relationship with the target content, where whether the association relationship exists between the target user and the target content is whether the target user clicks to view the interaction behavior of the target content; if yes, triggering the distribution unit;
the distribution unit is used for distributing the content labels to the target users.
9. The apparatus of claim 8, wherein if the first element is any one of the original features, the output features of the kth implicit feature layer comprise H k A determining unit configured to:
according to the first element and the H k A plurality of elements for determining the H in the importance information k Sub-importance information of the individual elements relative to the first element;
determining sub-importance information of the first element and the H k Intersection information of the individual elements;
and determining the output characteristics of the (k+1) th implicit characteristic layer according to the intersection information respectively corresponding to the elements included in the original characteristics.
10. The apparatus according to claim 8, wherein the determining unit is configured to:
Determining n output features corresponding to the n hidden feature layers through the user feature sub-model;
and determining the user characteristics corresponding to the user information according to the n output characteristics.
11. The apparatus according to claim 8, wherein the determining unit is configured to:
calculating similarity information between the user features and the tag features;
if the similarity information meets the association condition, determining that the target user can establish an association relationship with the target content;
and if the similarity information does not meet the association condition, determining that the target user cannot establish an association relationship with the target content.
12. The apparatus according to any one of claims 8-11, wherein the determining unit is configured to:
determining the label characteristics corresponding to the content labels through a label characteristic sub-model in the user prediction model;
and determining whether the target user can establish an association relationship with the target content according to the tag characteristics and the user characteristics through a predictor model in the user prediction model.
13. The apparatus of claim 12, further comprising a training unit;
The determining unit is further used for determining a training sample according to sample user information and sample content labels corresponding to the sample user; the sample content tag is used for identifying one type of sample content; the training sample comprises the sample user information, a sample content label and an identifier of whether the sample user can establish an association relationship with the sample content;
the training unit is used for training the user prediction model according to the training sample; in the training process, determining sample user characteristics corresponding to the sample user through the user characteristic sub-model, determining sample tag characteristics corresponding to the sample content tag through the tag characteristic sub-model, and correspondingly adjusting model parameters of the user prediction model if a result of whether the user determined through the prediction sub-model can establish an association relationship with the sample content is inconsistent with the identifier according to the sample user characteristics and the sample tag characteristics.
14. The apparatus according to any one of claims 8-11, wherein the acquisition unit is configured to:
acquiring the user information and the content tag in a first time period;
The determining unit is used for determining whether the target user can establish an association relationship with target content in a second time period according to the user characteristics and the tag characteristics; the first period of time is earlier than the second period of time.
15. An apparatus for user tag determination, the apparatus comprising a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to perform the method of any of claims 1-7 according to instructions in the program code.
16. A computer readable storage medium, characterized in that the computer readable storage medium is for storing a computer program for executing the method of any one of claims 1-7.
CN202011109112.2A 2020-10-16 2020-10-16 User tag determining method and related device Active CN112163164B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011109112.2A CN112163164B (en) 2020-10-16 2020-10-16 User tag determining method and related device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011109112.2A CN112163164B (en) 2020-10-16 2020-10-16 User tag determining method and related device

Publications (2)

Publication Number Publication Date
CN112163164A CN112163164A (en) 2021-01-01
CN112163164B true CN112163164B (en) 2024-03-15

Family

ID=73867517

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011109112.2A Active CN112163164B (en) 2020-10-16 2020-10-16 User tag determining method and related device

Country Status (1)

Country Link
CN (1) CN112163164B (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2016062445A (en) * 2014-09-19 2016-04-25 ヤフー株式会社 Label extraction device, label extraction method and program
CN110263265A (en) * 2019-04-10 2019-09-20 腾讯科技（深圳）有限公司 User tag generation method, device, storage medium and computer equipment
CN111046275A (en) * 2019-11-19 2020-04-21 腾讯科技（深圳）有限公司 User label determining method and device based on artificial intelligence and storage medium
CN111177569A (en) * 2020-01-07 2020-05-19 腾讯科技（深圳）有限公司 Recommendation processing method, device and equipment based on artificial intelligence
CN111625715A (en) * 2020-05-09 2020-09-04 北京达佳互联信息技术有限公司 Information extraction method and device, electronic equipment and storage medium
CN111666996A (en) * 2020-05-29 2020-09-15 湖北工业大学 High-precision equipment source identification method based on attention mechanism
CN111709789A (en) * 2020-06-18 2020-09-25 腾讯科技（深圳）有限公司 User conversion rate determining method and related equipment

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2016062445A (en) * 2014-09-19 2016-04-25 ヤフー株式会社 Label extraction device, label extraction method and program
CN110263265A (en) * 2019-04-10 2019-09-20 腾讯科技（深圳）有限公司 User tag generation method, device, storage medium and computer equipment
CN111046275A (en) * 2019-11-19 2020-04-21 腾讯科技（深圳）有限公司 User label determining method and device based on artificial intelligence and storage medium
CN111177569A (en) * 2020-01-07 2020-05-19 腾讯科技（深圳）有限公司 Recommendation processing method, device and equipment based on artificial intelligence
CN111625715A (en) * 2020-05-09 2020-09-04 北京达佳互联信息技术有限公司 Information extraction method and device, electronic equipment and storage medium
CN111666996A (en) * 2020-05-29 2020-09-15 湖北工业大学 High-precision equipment source identification method based on attention mechanism
CN111709789A (en) * 2020-06-18 2020-09-25 腾讯科技（深圳）有限公司 User conversion rate determining method and related equipment

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
融合用户标签和关系的微博用户相似性度量;吴树芳;徐建民;武晓波;;情报杂志;20141218(第12期);第170-173页 *

Also Published As

Publication number Publication date
CN112163164A (en) 2021-01-01

Similar Documents

Publication Publication Date Title
CN111813532B (en) Image management method and device based on multitask machine learning model
CN111125523B (en) Searching method, searching device, terminal equipment and storage medium
CN112203115B (en) Video identification method and related device
CN112749252A (en) Text matching method based on artificial intelligence and related device
CN112862021B (en) Content labeling method and related device
CN112163164B (en) User tag determining method and related device
CN110929882A (en) Feature vector calculation method based on artificial intelligence and related device
CN111723783B (en) Content identification method and related device
CN116259083A (en) Image quality recognition model determining method and related device
CN111988407B (en) Content pushing method and related device
CN116933149A (en) Object intention prediction method and device, electronic equipment and storage medium
CN116453005A (en) Video cover extraction method and related device
CN116450384A (en) Information processing method and related device
CN115080840A (en) Content pushing method and device and storage medium
CN113762587A (en) Material releasing method and device and related products
CN113569043A (en) Text category determination method and related device
CN111475732B (en) Information processing method and device
CN117056152B (en) Equipment detection method and related device
CN117115596B (en) Training method, device, equipment and medium of object action classification model
CN112685578B (en) Method and device for providing multimedia information content
CN116485468A (en) Information pushing method and related device
CN117216541A (en) Training method, prediction method and related products of material throwing effect prediction model
CN116992125A (en) Object recommendation method and device and storage medium
CN116992162A (en) Object recommendation method and device and storage medium
CN117112607A (en) Content searching method and related device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant