CN113220999A - User feature generation method and device, electronic equipment and storage medium - Google Patents

Abstract

The present disclosure provides a method, apparatus, electronic device and storage medium for generating user characteristics, and relates to the field of computer technology, in particular to the fields of artificial intelligence such as natural language processing and deep learning. The specific implementation scheme is: obtaining the first historical text data corresponding to the target user; analyzing the first historical text data to determine the first word segmentation set corresponding to the target user; The matching degree between each participle determines the number of participles under each topic contained in the first participle set; according to the number of participles under each topic contained in the first participle set, the user characteristics corresponding to the target user are determined. Thus, the user characteristics are determined based on the number of word segments of the target user under each topic, which improves the accuracy of the obtained user characteristics.

Description

Method, apparatus, electronic device and storage medium for generating user characteristics

technical field

The present disclosure relates to the field of computer technology, in particular to the fields of artificial intelligence such as natural language processing and deep learning, and in particular to a method and device for generating user characteristics, a method and device for training a model, electronic equipment and a storage medium.

Background technique

With the continuous development of Internet technology, many Internet-based products and services have emerged. In order to improve service quality and user experience, user characteristics can be analyzed, and personalized and accurate services can be provided to users based on user characteristics.

Therefore, how to improve the accuracy of the obtained user features is an urgent problem to be solved.

SUMMARY OF THE INVENTION

The present disclosure provides a method and apparatus for generating user characteristics, a method and apparatus for training a model, an electronic device and a storage medium.

According to an aspect of the present disclosure, a method for generating user characteristics is provided, including:

Obtain the first historical text data corresponding to the target user;

Parsing the first historical text data to determine the first segmented word set corresponding to the target user;

According to the degree of matching between each word segment in the first word segment set and each word segment corresponding to each topic, determine the number of word segments under each of the topics included in the first word segment set;

The user characteristic corresponding to the target user is determined according to the number of word segmentations under each topic contained in the first word segmentation set.

According to an aspect of the present disclosure, a model training method is provided, comprising:

Obtaining a training data set, wherein the training data set includes a plurality of historical text data corresponding to a plurality of users respectively;

Respectively analyze a plurality of historical text data corresponding to each of the users to determine the word segmentation set corresponding to each of the users;

Determine the word segmentation under each topic included in the word segmentation set corresponding to each of the users, and the labeling risk level corresponding to each of the users;

Input the word segmentation and the corresponding theme under each topic contained in the word segmentation set corresponding to each user to the initial neural network model to obtain the predicted risk level output by the initial neural network model;

According to the difference between the predicted risk level and the marked risk level, the initial neural network model is modified to generate a risk control model.

According to another aspect of the present disclosure, an apparatus for generating user characteristics is provided, comprising:

a first acquisition module, configured to acquire the first historical text data corresponding to the target user;

a first parsing module, configured to parse the first historical text data to determine the first segmented word set corresponding to the target user;

A first determination module, configured to determine the number of word segmentations under each of the topics contained in the first word segmentation set according to the degree of matching between each word segmentation in the first word segmentation set and each word segmentation corresponding to each theme;

The second determination module is configured to determine the user characteristics corresponding to the target user according to the number of word segmentations under each of the topics contained in the first word segmentation set.

According to another aspect of the present disclosure, a model training apparatus is provided, comprising:

a second acquisition module, configured to acquire a training data set, wherein the training data set includes a plurality of historical text data corresponding to a plurality of users respectively;

A second parsing module, configured to parse a plurality of historical text data corresponding to each of the users, respectively, to determine a word segmentation set corresponding to each of the users;

The eighth determination module is used to determine the word segmentation under each topic included in the word segmentation set corresponding to each of the users, and the labeling risk level corresponding to each of the users;

The second training module is used to input the word segmentation and the corresponding theme contained in the word segmentation set corresponding to each user to the initial neural network model to obtain the predicted risk level output by the initial neural network model; according to The difference between the predicted risk level and the marked risk level is used to modify the initial neural network model to generate a risk control model.

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

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of any of the above embodiments.

According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method according to any of the above embodiments.

According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the method according to any of the above embodiments.

It should be understood that what is described in this section is not intended to identify key or critical features of embodiments of the disclosure, nor is it intended to limit the scope of the disclosure. Other features of the present disclosure will become readily understood from the following description.

Description of drawings

The accompanying drawings are used for better understanding of the present solution, and do not constitute a limitation to the present disclosure. in:

FIG. 1 is a schematic flowchart of a method for generating user characteristics according to an embodiment of the present disclosure;

FIG. 2 is a schematic flowchart of another method for generating user characteristics according to an embodiment of the present disclosure;

FIG. 3 is a schematic flowchart of another method for generating user characteristics according to an embodiment of the present disclosure;

FIG. 4 is a schematic flowchart of another method for generating user characteristics according to an embodiment of the present disclosure;

FIG. 5 is a schematic flowchart of another method for generating user characteristics according to an embodiment of the present disclosure;

6 is a schematic flowchart of a model training method provided by an embodiment of the present disclosure;

7 is a schematic structural diagram of an apparatus for generating user characteristics according to an embodiment of the present disclosure;

FIG. 8 is a schematic structural diagram of a model training apparatus according to an embodiment of the present disclosure;

9 is a block diagram of an electronic device used to implement the method of an embodiment of the present disclosure.

Detailed ways

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding and should be considered as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted from the following description for clarity and conciseness.

The method, apparatus, electronic device, and storage medium for generating user characteristics according to the embodiments of the present disclosure are described below with reference to the accompanying drawings.

Artificial intelligence is a discipline that studies the use of computers to simulate certain thinking processes and intelligent behaviors of people (such as learning, reasoning, thinking, planning, etc.), both in the technical field of hardware and software. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies include computer vision technology, speech recognition technology, natural language processing technology, as well as deep learning, big data and other technologies. Processing technology, knowledge graph technology and other major directions.

NLP (Natural Language Processing, Natural Language Processing) is an important direction in the field of computer science and artificial intelligence. The content of NLP research includes but is not limited to the following branches: text classification, information extraction, automatic summarization, intelligent question answering, topic recommendation , machine translation, subject word recognition, knowledge base construction, deep text representation, named entity recognition, text generation, text analysis (lexical, syntactic, grammar, etc.), speech recognition and synthesis, etc.

Deep learning is a new research direction in the field of machine learning. Deep learning is to learn the inherent laws and representation levels of sample data, and the information obtained during these learning processes is of great help to the interpretation of data such as text, images, and sounds. Its ultimate goal is to enable machines to have the ability to analyze and learn like humans, and to recognize data such as words, images, and sounds.

FIG. 1 is a schematic flowchart of a method for generating a user feature according to an embodiment of the present disclosure.

As shown in Figure 1, the generation method of the user feature includes:

Step 101: Obtain first historical text data corresponding to the target user.

In the present disclosure, historical text data of a certain user within a preset time period in the past may be acquired. For convenience of distinction, the user may be referred to as a target user, and the acquired historical text data may be referred to as first historical text data. The first historical text data may be webpage content browsed by the target user, video content browsed by the target user, and the like.

In practical applications, historical text data within a preset time period before the operation time can also be acquired based on the time when the target user performs a certain operation. For example, if user A initiates a credit request at 16:00 on February 20, 2020, the historical text data of user A in the 15 days before the time can be obtained, such as the content of web pages browsed, online shopping, and video content browsed.

It should be noted that, in this disclosure, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations, and do not violate public order and good customs.

Step 102, parse the first historical text data to determine the first segmented word set corresponding to the target user.

In the present disclosure, the first historical text data can be parsed, such as removing interjections, auxiliary words, etc., and word segmentation and deduplication processing to obtain a plurality of word segmentations, which constitute a word segmentation set, which is called the first A set of words.

In practical applications, multiple historical text data of the target user may be obtained, and each historical text data can be parsed to obtain a word segmentation set of each historical text data. After that, according to the sequence of the generation time of the historical text data, the multiple word segmentation sets are combined and deduplicated to obtain the first word segmentation set. Alternatively, multiple historical text data can also be sorted in chronological order, and integrated into one historical text data, that is, the first historical text data can be obtained by integrating multiple historical text data, and then parsed to obtain multiple word segmentations. Form the first participle set.

Step 103: Determine the number of word segments under each topic included in the first word segment set according to the degree of matching between each word segment in the first word segment set and each word segment corresponding to each topic.

In the present disclosure, multiple topics and each word segment corresponding to each topic may be acquired in advance. The multiple topics obtained here and the respective word segments corresponding to each topic may be obtained by manually classifying the word segments in multiple documents.

For example, obtain two themes "game" and "travel", among which, the participles corresponding to the theme "game" are [King of Glory, jungler and glory assist the new heroes], and the participles corresponding to the theme "travel" are [attractions weather tickets Departure and destination] etc.

It should be noted that the word segmentation corresponding to the subject in the above examples is only an example, and cannot be regarded as a limitation of the present disclosure.

After obtaining the first word segmentation set, each word segment in the first word segmentation set may be matched with each word segment corresponding to each topic, so that each word segment in the first word segmentation set and each word segment corresponding to each topic can be matched The matching degree is to determine the word segmentation under each topic contained in the first word segmentation set, thereby determining the number of word segmentations under each topic included in the first word segmentation set.

When matching, the distance between the word vectors corresponding to the two participles can be calculated, and the matching degree of the two participles can be measured by the distance. Among them, the smaller the distance is, the higher the matching degree is, and the larger the distance is, the lower the matching degree is.

For example, match the participle p1 in the first participle set with each participle in topic a. If the matching degree between participle p1 and a participle in topic a is greater than the preset matching degree threshold, it can be considered that the first participle set contains the subtopics under the topic the participle.

Step 104: Determine the user characteristics corresponding to the target user according to the number of word segmentations under each topic contained in the first word segmentation set.

In the present disclosure, the number of word segmentations under each topic included in the first word segmentation set is greater than a preset number for comparison, and the topics with the number of word segmentations greater than the preset number are used as user characteristics corresponding to the target user. The preset number can be determined according to actual needs.

For example, the preset number is 0, and there are 8 topics in total. If there are 5 topics, the first word segmentation set contains the word segmentation under these topics, and these 5 topics can be used as the user characteristics corresponding to the target user.

Alternatively, one or more topics with the largest number of subtopic sub-words contained in the first word segmentation set may also be used as the corresponding user characteristics of the target user. For example, if the first word segmentation set contains the largest number of word segmentations under topic a and topic b, then topic a and topic b can be used as user characteristics corresponding to the target user.

Alternatively, the number of word segmentations under each topic included in the first word segmentation set may be directly used as the user feature corresponding to the target user.

For example, if there are 5 topics, and the number of segmentations under each topic contained in the first segmentation set is 6, 5, 4, 0, and 0, then the number of segmentations corresponding to the 5 topics can be the user characteristics corresponding to the target user. .

In the embodiment of the present disclosure, the first historical text data corresponding to the target user is analyzed to determine the first word segment set corresponding to the target user, and each word segment in the first word segment set is matched with each word segment under each theme, respectively, Determine the number of word segmentations under each topic included in the first word segmentation set, and determine the user characteristics corresponding to the target user according to the number of word segmentations under each topic included in the first word segmentation set. The number of word segmentations determines the user characteristics and improves the accuracy of the obtained user characteristics.

In an embodiment of the present disclosure, after the user characteristics are acquired, promotion information may also be pushed to the target user based on the user characteristics, thereby improving the accuracy of promotion information push. The following description will be made with reference to FIG. 2 , which is a schematic flowchart of another method for generating a user feature provided by an embodiment of the present disclosure.

As shown in Figure 2, the generation method of the user feature includes:

Step 201: Obtain first historical text data corresponding to the target user.

Step 202: Parse the first historical text data to determine the first segmented word set corresponding to the target user.

Step 203: Determine the number of word segments under each topic included in the first word segment set according to the degree of matching between each word segment in the first word segment set and each word segment corresponding to each topic.

Step 204: Determine user characteristics corresponding to the target user according to the number of word segmentations under each topic contained in the first word segmentation set.

In the present disclosure, steps 201 to 204 are similar to the above-mentioned steps 101 to 104, and thus are not repeated here.

Step 205: Determine the degree of association between the user characteristics and each piece of information to be promoted.

In the present disclosure, the degree of association between the user characteristics and each piece of information to be promoted can be calculated. The information to be promoted may be advertisements, videos, news, and the like.

For example, if the user features are games and travel, the degree of association between each piece of information to be promoted and the game, and the degree of association between each piece of information to be promoted and travel can be calculated.

Step 206: Determine target promotion information according to each correlation degree.

After obtaining the degree of association between the user feature and each piece of information to be promoted, the information to be promoted whose degree of association is greater than a preset threshold of the degree of association may be used as target promotion information.

It can be understood that the target promotion information may be one or multiple.

Step 207, push the target promotion information to the target user.

After acquiring the target promotion information, the target promotion information may be pushed to the target user through the client terminal used by the target user.

For example, if the user characteristic corresponding to the target user includes a game, then news, software, etc. related to the game may be recommended to the user.

In the embodiment of the present disclosure, after determining the user characteristics corresponding to the target users, the degree of association between the user characteristics and each piece of information to be promoted can also be determined, and according to each degree of association, the target promotion information is determined, and the target promotion is pushed to the target user. information. Therefore, the promotion information is pushed to the target user based on the user characteristics, so that the accuracy of the promotion information push can be improved.

In an embodiment of the present disclosure, the information to be promoted may also be pushed to the target user after obtaining the number of word segments under each topic included in the first word segment set. For example, if it is determined that the number of word segments under topic b contained in the first word segment set is the largest, target promotion information may be determined based on topic b, and the target promotion information may be pushed to target users.

In an embodiment of the present disclosure, after determining the number of word segmentations under each topic included in the first word segmentation set, it may also be determined whether the target user is a target risk user. The following description will be made with reference to FIG. 3 , which is a schematic flowchart of another method for generating a user feature provided by an embodiment of the present disclosure.

As shown in Figure 3, the generation method of the user feature includes:

Step 301: Obtain first historical text data corresponding to the target user.

Step 302: Parse the first historical text data to determine the first segmented word set corresponding to the target user.

In the present disclosure, steps 301 to 302 are similar to the above-mentioned steps 101 to 102, and thus are not repeated here.

Step 303: Determine the number of segmented words under the subject of the specified type contained in the first segmented set, where the subject of the specified type is related to the target risk.

After acquiring the number of word segmentations under each topic included in the first word segmentation set, the number of word segmentations under a topic of a specified type included in the first word segmentation set may be determined. Among them, the specified type topic is related to the target risk.

For example, if the specified type topic is installment payment, and the topic is related to the risk of advance consumption, the number of word segments under the installment payment topic included in the first word segment set can be determined. For another example, the target risk may also be the risk of overdue repayment.

It should be noted that the present disclosure does not limit the specified type of subject matter and target risk.

Step 304 , in the case that the number of word segmentations under the subject of the specified type contained in the first word segmentation set is greater than a preset threshold, determine the target user as a user with target risk.

In the present disclosure, if the number of word segmentations under the topic of the specified type contained in the first word segmentation set is greater than a preset threshold, it may be determined that the target user is a user with target risk.

For example, if the number of word segmentations under the installment theme included in the first word segmentation set is 20, which is greater than the preset threshold of 10, it can be considered that the target user has a risk of advanced consumption.

For another example, taking the credit risk control scenario as an example, if the number of word segmentations under the credit topic contained in the word segmentation set corresponding to a user is greater than the preset threshold, it can be considered that the user has the risk of overdue repayment, and the corresponding service can be refused to the user.

In the embodiment of the present disclosure, after determining the number of word segmentations under each topic included in the first word segmentation set, the number of word segmentations under a specified type of topic included in the first word segmentation set may also be determined, wherein the specified type of topic is related to the target risk, In the case that the number of word segments under the subject of the specified type contained in the first segment set is greater than a preset threshold, the target user is determined to be a user with a target risk. Therefore, whether the user is a target risk user can be identified according to the number of word segmentations under the subject of the specified type contained in the first word segmentation set, and it can be determined whether to provide certain services to the target user.

In an embodiment of the present disclosure, a clustering manner may be used to determine multiple topics and each word segment corresponding to the topic. The following description will be made with reference to FIG. 4 , which is a schematic flowchart of another method for generating a user feature provided by an embodiment of the present disclosure.

As shown in Figure 4, the method for generating the user feature may further include:

Step 401: Acquire a plurality of second historical text data corresponding to a plurality of users respectively.

Step 402 , respectively parse a plurality of second historical text data corresponding to each user to determine a second segmented word set corresponding to each user.

In the present disclosure, the manner of acquiring the second segmented word set is similar to the manner of acquiring the first segmented word set, and thus will not be repeated here.

Step 403: Clustering the word segments in the second word segment sets corresponding to the multiple users to obtain a plurality of thesaurus.

In the present disclosure, each user corresponds to a second segmented word set, and an LDA (Latent Dirichlet Allocation, latent Dirichlet Allocation) model can be used to cluster multiple second segmented word sets. After acquiring multiple second word segmentation sets, the number of topics obtained by clustering can be set, and the second word segmentation sets corresponding to each user are input into the initial LDA model to train the initial LDA model.

When the LDA model converges, the word segmentation probability distribution can be obtained, wherein the word segmentation probability distribution includes the probability that each word segment belongs to each topic, and then the word segmentation in multiple second word segmentation sets can be clustered according to the word segmentation probability distribution, for example, For each topic, a word segment with a probability greater than a preset probability can be used as the word segment corresponding to the topic, thereby obtaining multiple topic thesaurus. Among them, each subject thesaurus contains one or more participles.

For example, when training the LDA model, the number of topics is set to m, and based on the probability distribution of word segmentation when the model converges, a total of m topic thesaurus of topic_1, topic_2, ..., topic_m-1, and topic_m can be determined.

For example, topic_7:[The King's Glory, Jungle Glory, Auxiliary, and New Heroes]; topic_10:[Credit Card Application Card I Love Card Limit Love Card Installment Financial Repayment]; topic_11: [Scenic Spots, Weather, Train Tickets, Train Tickets, Bus Tickets, Airplane Tickets Departure and destination].

The above examples illustrate the word segmentation contained in the topic thesaurus topic_7, topic_10 and topic_11, or can also be considered as the word segmentation corresponding to the three topics of topic_7, topic_10 and topic_11 respectively. These word segmentations are part of the word segmentation in the topic thesaurus.

Step 404 , according to the matching degree between each segmented word in each thesaurus and the preset topic, determine the topics corresponding to each thesaurus respectively.

Because the above method is used to determine the topic corresponding to each topic thesaurus, for example, the topic thesaurus topic_11 is: [spots, weather, train ticket, bus ticket, plane, flight ticket departure destination], but I am not sure what the topic of topic_11 is. Therefore, in the present disclosure, the matching degree between each word segment in each thesaurus and the preset topic can be calculated, and each topic is determined according to the matching degree between the word segmentation and the preset topic in each thesaurus The thesaurus corresponds to the subject.

In the present disclosure, there are multiple preset topics, and the matching degree between each word segment in each thesaurus and each preset topic can be calculated. If each word segment in the thesaurus matches a certain preset topic are greater than the preset matching degree threshold, it can be determined that the theme corresponding to the thesaurus is the preset theme. In this way, the topic corresponding to each topic thesaurus can be determined, thereby obtaining multiple topics and the word segmentation corresponding to each topic. For example, it can be determined that the topic corresponding to the topic thesaurus topic_11 is "travel".

The method for generating user features in this embodiment of the present disclosure can be widely used in projects related to modeling and feature development. For example, based on a joint modeling project, credit data of multiple users can be obtained, such as including the user's borrowing time, etc. Based on the borrowing time, obtain the relevant historical text data of the user within a preset period of time before the borrowing time, such as browsing web content, video content, etc., and based on these historical text data, determine multiple thesaurus and the corresponding thesaurus for each thesaurus. theme.

In the embodiment of the present disclosure, it is also possible to obtain a plurality of second historical text data corresponding to a plurality of users, and parse the plurality of second historical text data to determine the second word segmentation sets corresponding to the plurality of users, and to analyze the second historical text data corresponding to the plurality of users. The word segments in the plurality of second word segmentation sets are clustered to obtain a plurality of thesaurus, and according to the matching degree between the word segments in each thesaurus and the preset topic, the respective themes corresponding to each thesaurus are determined. Thus, by using multiple second historical text data corresponding to multiple users, multiple subject thesaurus and themes corresponding to each subject thesaurus can be obtained, so that multiple subjects and the word segmentation corresponding to each subject can be used to determine User characteristics.

In one embodiment of the present disclosure, after determining the topics corresponding to each topic thesaurus, the risk control model can be obtained by training based on multiple second word segmentation sets corresponding to multiple users. The following description will be made with reference to FIG. 5 , which is a schematic flowchart of another method for generating a user feature provided by an embodiment of the present disclosure.

As shown in Figure 5, after determining the respective topics corresponding to each topic thesaurus, the above also includes:

Step 501 , according to the degree of matching between each segment in each second segment set and each segment corresponding to each topic, determine the segment under each topic included in each second segment set.

In the present disclosure, the method for determining the word segmentation under each topic included in the second word segmentation set is similar to the above-mentioned method for determining the word segmentation under each topic included in the first word segmentation set, so it is not repeated here.

Step 502: Determine the labeling risk level corresponding to each user based on the number of word segmentations under each topic included in each second word segmentation set.

In the present disclosure, the number of word segments under each topic included in each second segment set may be determined based on the number of segments under each topic included in each second segment set, so as to determine the label corresponding to each user Risk level.

Among them, the more the number of word segmentations under risk-related topics contained in the second word segmentation set, the higher the marked risk level.

For example, the topics related to overdue repayment risk include credit, installment payment, etc., according to the number of word segmentations under the subject of credit and installment payment included in the second word segmentation set corresponding to each user, to determine the corresponding labeling risk level for each user.

Step 503 , input the word segmentation and the corresponding topic under each topic contained in each second word segmentation set into the initial neural network model, so as to obtain the predicted risk level output by the initial neural network model.

In the present disclosure, the word segmentation and the corresponding topic under each topic included in the second word segmentation set can be input into the initial neural network model, and the initial neural network model can be used for prediction to obtain the predicted risk level corresponding to the user.

Step 504 , modifying the initial neural network model according to the difference between the predicted risk level and the marked risk level to generate a risk control model.

In the present disclosure, the difference between the predicted risk level and the marked risk level can be determined, and if the difference is greater than a preset threshold, the initial neural network model can be revised, and the revised model can be trained by using the remaining second segmentation set until the The model converges to generate the risk control model.

When training the risk control model, it can be trained through deep learning. Compared with other machine learning methods, deep learning performs better on large data sets.

In the present disclosure, the risk control model may be, for example, a credit risk control model. When determining the marked risk level, the number of word segments under the subject of credit and installment payment included in the second word segment set corresponding to each user may be used to determine the corresponding value of each user. , so as to train the credit risk control model. For another example, the risk control model may be an insurance risk control model. When determining the labeling risk level, the number of word segmentations under insurance-related topics included in the second word segmentation set corresponding to each user can be used to determine the number of words corresponding to each user. Label the risk level to train the insurance risk control model.

In the embodiment of the present disclosure, after each topic corresponding to each topic thesaurus is determined, each second word segment can also be determined according to the degree of matching between each word segment in each second word segment set and each word segment corresponding to each topic The word segmentation under each topic included in the set, and based on the number of word segmentations under each topic included in each second word segmentation set, determine the labeling risk level corresponding to each user, and classify each topic included in each second word segmentation set. The word segmentation and the corresponding topic below are input to the initial neural network model to obtain the predicted risk level output by the initial neural network model, and the initial neural network model is modified according to the difference between the predicted risk level and the marked risk level to generate wind control model. Thus, the word segmentation under each topic included in the second word segmentation set corresponding to each user can be used to train the risk control model.

In practical applications, after determining the topics corresponding to each topic thesaurus, the word segmentation under each topic contained in the second word segmentation set corresponding to each user can also be used to train a recommendation model, so as to use the recommendation model to target users. Push promotional information.

In an embodiment of the present disclosure, the risk control model may also be used to determine whether to respond to the user request of the target user.

In the present disclosure, when the user request sent by the target user is obtained, the first historical text data within the preset time period before the target user initiates the user request can be obtained, the first historical text data can be parsed, and the first word segmentation can be determined. Set, according to the degree of matching between each participle in the first participle set and each participle under each theme, determine the participle under each topic contained in the first participle set, and each part contained in the first participle set corresponding to the target user can be The word segmentation under each topic and the corresponding topic are input to the risk control model, and the risk control model outputs the risk level corresponding to the target user. If the risk level corresponding to the target user is lower than the preset risk level, it means that the risk of overdue repayment of the target user is relatively small, and the target user can respond to the user request of the target user. It can be understood that if the risk level corresponding to the target user is greater than or equal to the preset risk level, the user request of the target user can be rejected.

Alternatively, different user requests may be set to correspond to different risk levels. If the risk level corresponding to the target user is less than or equal to the risk level corresponding to the user request, the user request of the target user may be responded to.

For example, if the user's request is a credit request, and based on the credit risk control model, it is determined that the user's corresponding risk level is lower than the preset risk level, and the user's credit request can be responded to, so that the corresponding service can be provided to the user.

In the embodiment of the present disclosure, when the user request sent by the target user is obtained, the word segmentation and the corresponding topic under each topic contained in the first word segmentation set corresponding to the target user can be input into the risk control model to determine the target The risk level corresponding to the user; if the risk level is less than the preset risk level, respond to the user request. Therefore, based on the word segmentation under each topic contained in the first word segmentation set corresponding to the target user, the risk control model can be used to determine whether to respond to the user request of the target user, which improves the model's distinguishing effect on overdue users and reduces economic losses. .

In order to implement the above embodiments, the embodiments of the present disclosure further provide a model training method. FIG. 6 is a schematic flowchart of a model training method provided by an embodiment of the present disclosure.

As shown in Figure 6, the model training method includes:

Step 601: Acquire a training data set, wherein the training data set includes a plurality of historical text data corresponding to a plurality of users respectively.

In the present disclosure, a plurality of historical text data corresponding to a plurality of users, such as browsed webpage content, browsed video content, etc., can be obtained as a training data set.

It should be noted that, in this disclosure, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations, and do not violate public order and good customs.

Step 602, respectively parse a plurality of historical text data corresponding to each user to determine a word segmentation set corresponding to each user.

In the present disclosure, each of the multiple users corresponds to multiple historical text data, and each historical text data can be parsed, such as word segmentation and deduplication processing, to obtain a word segmentation set of each historical text data, and then According to the sequence of the generation time of the historical text data, the multiple word segmentation sets are combined and deduplicated to obtain the word segmentation set corresponding to the user.

Alternatively, when analyzing multiple historical text data corresponding to each user, the multiple historical text data corresponding to the user may be sorted in chronological order and integrated into one historical text data, that is, the first historical text data may be multiple It is obtained by integrating the historical text data, and then parsed to obtain multiple word segmentations, which constitute a word segmentation set.

Step 603: Determine the word segmentation under each topic included in the word segmentation set corresponding to each user, and the labeling risk level corresponding to each user.

Step 604 , input the word segmentation and the corresponding topic under each topic contained in the word segmentation set corresponding to each user into the initial neural network model to obtain the predicted risk level output by the initial neural network model.

Step 605, modifying the initial neural network model according to the difference between the predicted risk level and the marked risk level to generate a risk control model.

In the present disclosure, steps 603-605 are similar to the above-mentioned steps 502-504, so they will not be repeated here.

In the present disclosure, the risk control model may be, for example, a credit risk control model. When determining the labeling risk level, the labeling corresponding to each user may be determined according to the number of word segmentations under the subject of credit and installment payment included in the word segmentation set corresponding to each user. risk level, so as to train the credit risk control model. For another example, the risk control model may be an insurance risk control model. When determining the labeling risk level, the labeling risk corresponding to each user may be determined according to the number of word segmentations under insurance-related topics contained in the word segmentation set corresponding to each user. level, so as to train the insurance risk control model.

In the embodiment of the present disclosure, by acquiring training data, the plurality of historical text data corresponding to each user in the training data set are respectively parsed to determine the word segmentation set corresponding to each user, and the word segmentation set corresponding to each user is determined. The word segmentation under each topic, as well as the labeling risk level corresponding to each user, input the word segmentation and corresponding topic under each topic contained in the word segmentation set corresponding to each user to the initial neural network model to obtain the output of the initial neural network model. Predict the risk level, and modify the initial neural network model according to the difference between the predicted risk level and the marked risk level to generate a risk control model. Therefore, by obtaining the word segmentation set corresponding to each user, determining the word segmentation under each topic contained in the word segmentation set corresponding to each user, and using the word segmentation under each topic contained in the word segmentation set corresponding to each user, the training result can be obtained. Risk control model, thereby improving the accuracy of the model.

In an embodiment of the present disclosure, when determining the word segmentation under each topic included in the word segmentation set corresponding to each user, a document topic generation model, such as an LDA model, can be used to perform the segmentation on multiple word segmentation sets corresponding to multiple users. clustering.

After obtaining multiple word segmentation sets, the number of topics obtained by clustering can be set, and each word segmentation in the word segmentation set corresponding to each user is input into the initial document topic generation model to train the initial document topic generation model.

When the document topic generation model converges, the word segmentation probability distribution can be obtained, wherein the word segmentation probability distribution includes the probability that each word segment belongs to each topic, and then the word segmentation in multiple word segmentation sets can be clustered according to the word segmentation probability distribution. For example, for For each topic, the word segmentation with a probability greater than the preset probability can be used as the word segmentation corresponding to the topic, so that the word segmentation corresponding to each topic can be obtained. thesaurus. Among them, the number of subject thesaurus is the same as the number of subjects, and each subject thesaurus contains one or more word segmentations.

When acquiring multiple subject thesaurus, by using the word segmentation probability distribution obtained by the document subject generation model, multiple subject thesaurus are obtained, which improves the accuracy of the subject thesaurus.

Since multiple thesaurus can be obtained by using the document topic generation model, but it is not sure what the specific topic corresponding to each thesaurus is, therefore, in the present disclosure, each segmented word and preset topic in each thesaurus can be calculated. The matching degree between each thesaurus is determined according to the matching degree between each segmented word in each thesaurus and the preset topic, and the topic corresponding to each thesaurus is determined.

In the present disclosure, the method of determining the subject corresponding to each subject thesaurus is similar to the above-mentioned step 404, so it is not repeated here.

After the topic corresponding to each thesaurus is determined, the word segmentation under each topic included in the word segmentation set corresponding to each user can be similar to the above-mentioned method of determining the word segmentation under each topic included in the first word segmentation set. Therefore, in This will not be repeated here.

When marking the risk level corresponding to each user, the number of word segmentations under each topic included in the word segmentation set corresponding to each user can be determined according to the number of word segmentations under each topic contained in the word segmentation set corresponding to each user. The word segmentation set corresponding to each user includes the number of word segmentations under risk-related topics, and the labeling risk level corresponding to each user can be determined.

In the present disclosure, the labeling risk level corresponding to each user is determined based on the number of word segments under each topic included in the word segment set corresponding to each user, thereby improving the labeling accuracy.

In order to implement the above embodiments, an embodiment of the present disclosure further provides an apparatus for generating user characteristics. FIG. 7 is a schematic structural diagram of an apparatus for generating user characteristics according to an embodiment of the present disclosure.

As shown in FIG. 7 , the apparatus 700 for generating user characteristics includes: a first acquiring module 710 , a first parsing module 720 , a first determining module 730 and a second determining module 740 .

a first obtaining module 710, configured to obtain the first historical text data corresponding to the target user;

A first parsing module 720, configured to parse the first historical text data to determine a first segmented word set corresponding to the target user;

The first determination module 730 is configured to determine the number of word segmentations under each of the topics contained in the first word segmentation set according to the degree of matching between each word segmentation in the first word segmentation set and each word segmentation corresponding to each topic ;

The second determination module 740 is configured to determine the user characteristic corresponding to the target user according to the number of word segmentations under each topic included in the first word segmentation set.

In a possible implementation manner of the embodiment of the present disclosure, the apparatus may further include:

a third determining module, configured to determine each correlation degree between the user feature and each information to be promoted;

a fourth determining module, configured to determine target promotion information according to the respective degrees of association;

A push module, configured to push the target promotion information to the target user.

In a possible implementation manner of the embodiment of the present disclosure, the first determination module 730 is further configured to determine the number of word segments under a specified type of topic included in the first word segmentation set, where the specified type of topic is related to a target risk ;

The second determination module 740 is further configured to determine that the target user is a user with the target risk when the number of word segmentations under the subject of the specified type contained in the first segmentation set is greater than a preset threshold.

In a possible implementation manner of the embodiment of the present disclosure, the first obtaining module 710 is further configured to obtain a plurality of second historical text data corresponding to a plurality of users respectively;

The first parsing module 720 is further configured to parse a plurality of second historical text data corresponding to each of the users to determine a second set of words corresponding to each of the users;

The device may also include:

a clustering module, configured to cluster the word segments in a plurality of second word segmentation sets corresponding to the plurality of users to obtain a plurality of thesaurus;

The fifth determining module is configured to determine the topics corresponding to each thesaurus according to the degree of matching between each segmented word in each of the thesaurus and the preset topic.

In a possible implementation manner of the embodiment of the present disclosure, the first determining module 730 is further configured to, according to the degree of matching between each participle in each of the second participle sets and each participle corresponding to each topic, respectively, determining a participle under each of the topics contained in each of the second participle sets;

The device may also include:

A sixth determination module, configured to determine the labeling risk level corresponding to each of the users based on the number of word segmentations under each of the topics contained in each of the second word segmentation sets;

The first training module is used to input the word segmentation and the corresponding topic under each of the topics contained in each of the second word segmentation sets to the initial neural network model to obtain the predicted risk level output by the initial neural network model ; According to the difference between the predicted risk level and the marked risk level, modify the initial neural network model to generate a risk control model.

In a possible implementation manner of the embodiment of the present disclosure, the apparatus may further include:

The seventh determination module is used to input the word segmentation and the corresponding theme under each of the topics contained in the first word segmentation set corresponding to the target user in the case of obtaining the user request sent by the target user into the the risk control model to determine the risk level corresponding to the target user;

A response module, configured to respond to the user request when the risk level is less than a preset risk level.

It should be noted that, the explanations of the foregoing embodiments of the method for generating user characteristics are also applicable to the device for generating user characteristics in this embodiment, so they are not repeated here.

In the embodiment of the present disclosure, the first historical text data corresponding to the target user is analyzed to determine the first word segment set corresponding to the target user, and each word segment in the first word segment set is matched with each word segment under each theme, respectively, Determine the number of word segmentations under each topic included in the first word segmentation set, and determine the user characteristics corresponding to the target user according to the number of word segmentations under each topic included in the first word segmentation set. The number of word segmentations determines the user characteristics and improves the accuracy of the obtained user characteristics.

In order to realize the above embodiments, the present disclosure also proposes a model training device. FIG. 8 is a schematic structural diagram of a model training apparatus according to an embodiment of the present disclosure.

As shown in FIG. 8 , the model training apparatus includes: a second acquisition module 810 , a second analysis module 820 , an eighth determination module 830 and a second training module 840 .

The second obtaining module 810 is configured to obtain a training data set, wherein the training data set includes a plurality of historical text data corresponding to a plurality of users respectively;

A second parsing module 820, configured to parse a plurality of historical text data corresponding to each of the users, respectively, to determine a word segmentation set corresponding to each of the users;

The eighth determination module 830 is used to determine the word segmentation under each topic included in the word segmentation set corresponding to each of the users, and the labeling risk level corresponding to each of the users;

The second training module 840 is used to input the word segmentation and the corresponding topic under each topic contained in the word segmentation set corresponding to each user to the initial neural network model, to obtain the predicted risk level output by the initial neural network model; According to the difference between the predicted risk level and the marked risk level, the initial neural network model is modified to generate a risk control model.

In a possible implementation manner of the embodiment of the present disclosure, the eighth determination module 830 includes:

a clustering unit, configured to cluster the word segments in the multiple word segment sets corresponding to the multiple users to obtain multiple subject thesaurus;

a first determining unit, configured to determine the respectively corresponding themes of each thesaurus according to the matching degree between each word segment in each of the thesaurus and a preset theme;

The second determination unit is used to determine the word segmentation under each of the described topics contained in each of the described word segmentation sets according to the degree of matching between the respective word segmentations in each of the described word segmentation sets and the respective word segmentations corresponding to the respective topics;

The third determining unit is configured to determine the labeling risk level corresponding to each user based on the number of word segments under each topic included in the word segment set corresponding to each user.

In a possible implementation manner of the embodiment of the present disclosure, the clustering unit is used for:

Inputting each word segment in the plurality of word segmentation sets into a document topic generation model to obtain a word segmentation probability distribution, wherein the word segmentation probability distribution includes the probability that each word segment belongs to each topic;

The word segmentations in the multiple word segmentation sets are clustered according to the word segmentation probability distribution to obtain multiple subject thesaurus.

In the embodiment of the present disclosure, by acquiring training data, the plurality of historical text data corresponding to each user in the training data set are respectively parsed to determine the word segmentation set corresponding to each user, and the word segmentation set corresponding to each user is determined. The word segmentation under each topic, as well as the labeling risk level corresponding to each user, input the word segmentation and corresponding topic under each topic contained in the word segmentation set corresponding to each user to the initial neural network model to obtain the output of the initial neural network model. Predict the risk level, and modify the initial neural network model according to the difference between the predicted risk level and the marked risk level to generate a risk control model. Therefore, by obtaining the word segmentation set corresponding to each user, determining the word segmentation under each topic contained in the word segmentation set corresponding to each user, and using the word segmentation under each topic contained in the word segmentation set corresponding to each user, the training result can be obtained. Risk control model, thereby improving the accuracy of the model.

According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

FIG. 9 shows a schematic block diagram of an example electronic device 900 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.

As shown in FIG. 9, the device 900 includes a computing unit 901, which can be loaded into a RAM (Random Access Memory, random access/access) according to a computer program stored in a ROM (Read-Only Memory) 902 or from a storage unit 908 memory) 903 to perform various appropriate actions and processes. In the RAM 903, various programs and data necessary for the operation of the device 900 can also be stored. The computing unit 901 , the ROM 902 , and the RAM 903 are connected to each other through a bus 904 . An I/O (Input/Output) interface 905 is also connected to the bus 904 .

Various components in the device 900 are connected to the I/O interface 905, including: an input unit 906, such as a keyboard, mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a magnetic disk, an optical disk, etc. ; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 909 allows the device 900 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks.

Computing unit 901 may be various general-purpose and/or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a CPU (Central Processing Unit, central processing unit), a GPU (Graphic Processing Units, graphics processing unit), various dedicated AI (Artificial Intelligence, artificial intelligence) computing chips, various operating The computing unit of the machine learning model algorithm, DSP (Digital Signal Processor, digital signal processor), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as the generation method of the user characteristics. For example, in some embodiments, the method of generating a user profile may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 908 . In some embodiments, part or all of the computer program may be loaded and/or installed on device 900 via ROM 902 and/or communication unit 909 . When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the method of generating the user profile described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured by any other suitable means (eg, by means of firmware) to perform the method of generating user characteristics.

Various implementations of the systems and techniques described herein above may be implemented in digital electronic circuitry, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits) , ASSP (Application Specific Standard Product), SOC (System On Chip, System On Chip), CPLD (Complex Programmable Logic Device), computer hardware, firmware, software, and/or their implemented in combination. These various embodiments may include being implemented in one or more computer programs executable and/or interpretable on a programmable system including at least one programmable processor that The processor, which may be a special purpose or general-purpose programmable processor, may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device an output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer or other programmable data processing apparatus, such that the program code, when executed by the processor or controller, performs the functions/functions specified in the flowcharts and/or block diagrams. Action is implemented. The program code may execute entirely on the machine, partly on the machine, partly on the machine and partly on a remote machine as a stand-alone software package or entirely on the remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in connection with the instruction execution system, apparatus or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only-Memory) Or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage device, magnetic storage device, or any suitable combination of the above.

To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having: a display device (eg, a CRT (Cathode-Ray Tube) or an LCD (Cathode-Ray Tube) for displaying information to the user Liquid Crystal Display (liquid crystal display) monitor); and a keyboard and pointing device (eg, mouse or trackball) through which a user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (eg, visual feedback, auditory feedback, or tactile feedback); and can be in any form (including acoustic input, voice input, or tactile input) to receive input from the user.

The systems and techniques described herein may be implemented on a computing system that includes back-end components (eg, as a data server), or a computing system that includes middleware components (eg, an application server), or a computing system that includes front-end components (eg, a user's computer having a graphical user interface or web browser through which a user may interact with implementations of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system. The components of the system may be interconnected by any form or medium of digital data communication (eg, a communication network). Examples of communication networks include: LAN (Local Area Network, Local Area Network), WAN (Wide Area Network, Wide Area Network), Internet, and blockchain networks.

A computer system can include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also known as a cloud computing server or a cloud host. It is a host product in the cloud computing service system to solve the existing management problems in traditional physical hosts and VPS services (VirtualPrivate Server, virtual private server). Difficulty, weak business expansion defects. The server can also be a server of a distributed system, or a server combined with a blockchain.

It should be noted that the above-mentioned electronic device can also execute the model training method in the present disclosure.

According to an embodiment of the present disclosure, the present disclosure also provides a computer program product that, when executed by an instruction processor in the computer program product, executes the user feature generation method or the model training method proposed by the above embodiments of the present disclosure.

It should be understood that steps may be reordered, added or deleted using the various forms of flow shown above. For example, the steps described in the present disclosure can be executed in parallel, sequentially, or in different orders. As long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, there is no limitation herein.

The above-mentioned specific embodiments do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure should be included within the protection scope of the present disclosure.

Claims (21)

1. A method for generating user characteristics, comprising: Obtain the first historical text data corresponding to the target user; Parsing the first historical text data to determine the first segmented word set corresponding to the target user; According to the degree of matching between each word segment in the first word segment set and each word segment corresponding to each topic, determine the number of word segments under each of the topics included in the first word segment set; The user characteristic corresponding to the target user is determined according to the number of word segmentations under each topic contained in the first word segmentation set. 2. The method according to claim 1, wherein after the determining the user characteristic corresponding to the target user, the method further comprises: determining the degree of association between the user characteristics and each piece of information to be promoted; According to each correlation degree, determine the target promotion information; Pushing the target promotion information to the target user. 3. The method of claim 1, wherein the determining the number of word segments under each of the topics included in the first word segment set comprises: Determining the number of segmented words under a specified type of subject included in the first segmented set, wherein the specified type of subject is related to the target risk; Determining the user characteristics corresponding to the target user according to the number of word segmentations under each of the topics contained in the first word segmentation set, including: In the case that the number of segmented words under the subject of the specified type contained in the first segmented word set is greater than a preset threshold, it is determined that the target user is a user with the target risk. 4. The method of claim 1, further comprising: acquiring multiple second historical text data corresponding to multiple users respectively; Analyzing a plurality of second historical text data corresponding to each of the users respectively, to determine the second segmented word set corresponding to each of the users; Clustering the word segments in a plurality of second word segmentation sets corresponding to the plurality of users to obtain a plurality of thesaurus; According to the matching degree between each segmented word in each of the thesaurus and the preset topic, the topics corresponding to each thesaurus are determined. 5. The method according to claim 4, wherein, after said determining the themes respectively corresponding to each thesaurus, further comprising: According to the degree of matching between each participle in each of the second participle sets and each participle corresponding to each topic, determine the participle under each of the topics contained in each of the second participle sets; Determine the labeling risk level corresponding to each of the users based on the number of word segmentations under each of the topics contained in each of the second word segmentation sets; Inputting the word segmentation and the corresponding topic under each of the topics contained in each of the second word segmentation sets to the initial neural network model to obtain the predicted risk level output by the initial neural network model; According to the difference between the predicted risk level and the marked risk level, the initial neural network model is modified to generate a risk control model. 6. The method of claim 5, wherein the method further comprises: In the case of obtaining the user request sent by the target user, input the word segmentation and the corresponding topic under each topic contained in the first word segmentation set corresponding to the target user into the risk control model, so as to determining the risk level corresponding to the target user; In the case that the risk level is less than a preset risk level, responding to the user request. 7. A model training method, comprising: Obtaining a training data set, wherein the training data set includes a plurality of historical text data corresponding to a plurality of users respectively; Respectively analyze a plurality of historical text data corresponding to each of the users to determine the word segmentation set corresponding to each of the users; Determine the word segmentation under each topic included in the word segmentation set corresponding to each of the users, and the labeling risk level corresponding to each of the users; Input the word segmentation and the corresponding theme under each topic contained in the word segmentation set corresponding to each user to the initial neural network model to obtain the predicted risk level output by the initial neural network model; According to the difference between the predicted risk level and the marked risk level, the initial neural network model is modified to generate a risk control model. 8. The method according to claim 7, wherein the determining the word segmentation under each topic contained in the word segmentation set corresponding to each of the users, and the labeling risk level corresponding to each of the users, comprises: Clustering the word segmentations in the multiple word segmentation sets corresponding to the multiple users to obtain multiple subject thesaurus; According to the matching degree between each segmented word in each of the thesaurus and the preset topic, determine the topics corresponding to each thesaurus respectively; According to the matching degree between each word segment in each described word segment set and each word segment corresponding to each topic respectively, determine the word segment under each described topic included in each described word segment set; Based on the number of word segmentations under each of the topics included in the word segmentation set corresponding to each of the users, a labeling risk level corresponding to each of the users is determined. 9. The method according to claim 8, wherein the clustering of the word segmentations in the plurality of word segmentation sets corresponding to the plurality of users to obtain a plurality of thesaurus, comprising: Inputting each word segment in the plurality of word segmentation sets into a document topic generation model to obtain a word segmentation probability distribution, wherein the word segmentation probability distribution includes the probability that each word segment belongs to each topic; The word segmentations in the multiple word segmentation sets are clustered according to the word segmentation probability distribution to obtain multiple subject thesaurus. 10. A device for generating user characteristics, comprising: a first acquisition module, configured to acquire the first historical text data corresponding to the target user; a first parsing module, configured to parse the first historical text data to determine the first segmented word set corresponding to the target user; A first determination module, configured to determine the number of word segmentations under each of the topics contained in the first word segmentation set according to the degree of matching between each word segmentation in the first word segmentation set and each word segmentation corresponding to each theme; The second determination module is configured to determine the user characteristics corresponding to the target user according to the number of word segmentations under each of the topics contained in the first word segmentation set. 11. The apparatus of claim 10, wherein the apparatus further comprises: a third determining module, configured to determine the degree of association between the user characteristics and each piece of information to be promoted; a fourth determining module, used for determining target promotion information according to each correlation degree; A push module, configured to push the target promotion information to the target user. 12. The apparatus according to claim 10, wherein the first determining module is further configured to determine the number of word segments under a specified type of topic included in the first word segmentation set, wherein the specified type of topic and target risk related; The second determination module is further configured to determine that the target user is a user with the target risk in the case that the number of word segmentations under the subject of the specified type contained in the first word segmentation set is greater than a preset threshold. 13. The apparatus according to claim 10, wherein the first obtaining module is further configured to obtain a plurality of second historical text data corresponding to a plurality of users respectively; The first parsing module is further configured to parse a plurality of second historical text data corresponding to each of the users, respectively, to determine a second set of words corresponding to each of the users; The device also includes: a clustering module, configured to cluster the word segments in a plurality of second word segmentation sets corresponding to the plurality of users to obtain a plurality of thesaurus; The fifth determining module is configured to determine the topics corresponding to each thesaurus according to the degree of matching between each segmented word in each of the thesaurus and the preset topic. 14. The apparatus according to claim 13, wherein the first determining module is further configured to determine according to the degree of matching between each participle in each of the second participle sets and each participle corresponding to each topic respectively, to determine a participle under each of the topics contained in each of the second participle sets; The device also includes: A sixth determination module, configured to determine the labeling risk level corresponding to each of the users based on the number of word segmentations under each of the topics contained in each of the second word segmentation sets; The first training module is used to input the word segmentation and the corresponding topic under each of the topics contained in each of the second word segmentation sets to the initial neural network model to obtain the predicted risk level output by the initial neural network model ; According to the difference between the predicted risk level and the marked risk level, modify the initial neural network model to generate a risk control model. 15. The apparatus of claim 14, wherein the apparatus further comprises: The seventh determination module is used to input the word segmentation and the corresponding theme under each of the topics contained in the first word segmentation set corresponding to the target user in the case of obtaining the user request sent by the target user into the the risk control model to determine the risk level corresponding to the target user; A response module, configured to respond to the user request when the risk level is less than a preset risk level. 16. A model training device, comprising: a second acquisition module, configured to acquire a training data set, wherein the training data set includes a plurality of historical text data corresponding to a plurality of users respectively; A second parsing module, configured to parse a plurality of historical text data corresponding to each of the users, respectively, to determine a word segmentation set corresponding to each of the users; The eighth determination module is used to determine the word segmentation under each topic included in the word segmentation set corresponding to each of the users, and the labeling risk level corresponding to each of the users; The second training module is used to input the word segmentation and the corresponding theme contained in the word segmentation set corresponding to each user to the initial neural network model to obtain the predicted risk level output by the initial neural network model; according to The difference between the predicted risk level and the marked risk level is used to modify the initial neural network model to generate a risk control model. 17. The apparatus of claim 16, wherein the eighth determining module comprises: a clustering unit, configured to cluster the word segments in the multiple word segment sets corresponding to the multiple users to obtain multiple subject thesaurus; a first determining unit, configured to determine the respectively corresponding themes of each thesaurus according to the matching degree between each word segment in each of the thesaurus and a preset theme; The second determination unit is used to determine the word segmentation under each of the described topics contained in each of the described word segmentation sets according to the degree of matching between the respective word segmentations in each of the described word segmentation sets and the respective word segmentations corresponding to the respective topics; The third determining unit is configured to determine the labeling risk level corresponding to each user based on the number of word segments under each topic included in the word segment set corresponding to each user. 18. The apparatus of claim 17, wherein the clustering unit is used for: Inputting each word segment in the plurality of word segmentation sets into a document topic generation model to obtain a word segmentation probability distribution, wherein the word segmentation probability distribution includes the probability that each word segment belongs to each topic; The word segmentations in the plurality of word segmentation sets are clustered according to the word segmentation probability distribution to obtain a plurality of subject thesaurus. 19. An electronic device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the execution of any of claims 1-9 Methods. 20. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any of claims 1-9. 21. A computer program product comprising a computer program which, when executed by a processor, implements the method of any of claims 1-9.

Images