Method, system and equipment for constructing user tag rule engine and storage medium

Abstract

The embodiment of the invention discloses a method, a system and equipment for constructing a user tag rule engine and a storage medium, and relates to the technical field of financial data processing, and the method comprises the following steps: carrying out word segmentation processing on historical associated data of a historical user to obtain a plurality of segmented words; determining a plurality of segmented word groups based on the similarity among the plurality of segmented words to form a first word bank; according to the first word bank, a plurality of search rules corresponding to the segmented word groups are generated, and different segmented word groups correspond to different search rules. According to a plurality of search rules, a user tag rule engine is constructed, the user tag rule engine is used for determining the user tag of the user, when complex and diversified user data such as high-similarity word segmentation in banking business is faced, the number of target rules corresponding to each word segmentation expanded by compiling new codes can be reduced, and the efficiency of the banking business is improved. Therefore, the situation of continuously changing business requirements is adapted, and the development time cost and the labor cost in the process of determining the user tag are reduced.

Description

Technical Field

[0001] The present application relates to the technical field of financial data processing, and in particular to a method, system, device and storage medium for building a user tag rule engine. Background Art

[0002] The rule engine evolved from the inference engine. It is a component embedded in the application. It separates business decisions from the application code and uses predefined semantic modules to write business decisions. It can receive data input, interpret business rules, and make business decisions based on business rules. The application of rule engines is particularly significant in banking business. Taking user tag management as an example, banks can use rule engines to analyze user data and convert them into user tags. These user tags not only help banks better understand user needs, but also provide decision support for personalized services and precision marketing.

[0003] In traditional business systems, rule engines usually rely on a set of standard rules that are set through configuration pages. When faced with complex and diverse user data in banking business, it is often necessary to constantly write new code to expand the rule set to adapt to changing business needs. This approach not only increases the time cost of development, but also increases the consumption of human resources. However, there is currently no better solution.

[0004] Application Contents

[0005] In view of this, one of the purposes of the present application is to provide a method, system, device and storage medium for constructing a user tag rule engine, which can construct a user tag rule engine for determining user tags, reduce the development time cost when facing complex and diverse user data, and reduce the consumption of human resources.

[0006] To achieve the above purpose, the technical solution of this application is implemented as follows:

[0007] In a first aspect, an embodiment of the present application provides a method for building a user tag rule engine, including:

[0008] Perform word segmentation on the historical associated data of historical users to obtain multiple word segments;

[0009] Based on the similarities between the multiple participles, determine multiple participle groups from the multiple participles to form a first word library including the multiple participle groups, wherein each participle group includes at least two participles whose similarities meet a preset similarity condition;

[0010] According to the first word library, a plurality of search rules corresponding to the plurality of word groups are generated, wherein different word groups correspond to different search rules, and each search rule is used to determine a user tag of any word in the word group corresponding to it;

[0011] A user tag rule engine is constructed according to multiple search rules, and the user tag rule engine is used to determine the user tag of the user.

[0012] In a possible implementation, the historical associated data of the historical users is segmented to obtain multiple segmented words, including:

[0013] Obtain user portrait data of historical users;

[0014] Based on user portrait data, target word segmentation rules for historical users are formed;

[0015] The historical associated data of the historical user is segmented according to the target segmentation rule of the historical user to obtain multiple segmentations.

[0016] In a possible implementation, before generating a plurality of search rules corresponding to a plurality of word groups according to the first word library, the method further includes:

[0017] Based on the word frequency of each segmented word, determine multiple target segmented words from the multiple segmented words to form a second word library including the multiple target segmented words;

[0018] According to the first word library, a plurality of search rules corresponding to the plurality of word groups are generated, including:

[0019] Generate a plurality of first search rules corresponding to a plurality of word groups according to the first word library;

[0020] Generating a plurality of second search rules corresponding to a plurality of target word segments according to the second word library;

[0021] The multiple search rules include multiple first search rules and multiple second search rules.

[0022] In a possible implementation, a plurality of first search rules corresponding to a plurality of word groups are generated according to the first word library, including:

[0023] The first word library is analyzed by using a preset deep learning model to obtain feature data of each word segment in the first word library, wherein the feature data of each word segment includes semantic information and contextual relationship information of each word segment;

[0024] Generate a plurality of first search rules according to the semantic information and contextual relationship information of each word segment;

[0025] According to the second word library, a plurality of second search rules corresponding to the plurality of second target word segments are generated, including:

[0026] The second word library is analyzed by using a preset deep learning model to obtain feature data of each target word segment in the second word library, wherein the feature data of each target word segment includes semantic information and contextual relationship information of each word segment;

[0027] Generate multiple second search rules according to the semantic information and contextual relationship information of each target word segment;

[0028] Among them, the preset deep learning models include a recurrent neural network model, a transformer model, a natural language processing model and a recurrent neural network model.

[0029] In a possible implementation, based on the word frequency of each segmented word, multiple target segmented words are determined from multiple segmented words to form a second word library including the multiple target segmented words, including:

[0030] According to the frequency of each participle, the weight value of each participle is determined, wherein the frequency of the word is directly proportional to the weight value;

[0031] When the source channels of each segmented word are determined, the segmented words with the same source channel are assigned to the same data group to form an initial data set, wherein the initial data set includes at least one data group;

[0032] Get the co-occurrence frequency between different word segments in each data group;

[0033] Determine a first sub-participle in each data group, where the first sub-participle is a participle corresponding to a co-occurrence frequency greater than or equal to a preset frequency;

[0034] The first sub-participle in each data group whose weight value is greater than or equal to the preset weight value is used as the second sub-participle;

[0035] All second sub-words in all data groups are used as multiple target word segments to form a second word library including the multiple target word segments.

[0036] In a possible implementation, based on the similarity between the multiple segmented words, determining multiple segmented word groups from the multiple segmented words includes:

[0037] determining a plurality of preset similarity threshold ranges;

[0038] Based on the similarity between the multiple segmented words, the segmented words corresponding to the similarity within each preset similarity threshold range are respectively formed into a segmented word group to obtain multiple segmented word groups.

[0039] In a possible implementation, after constructing a user tag rule engine according to multiple search rules, the method further includes:

[0040] When target-related data of a newly added user is detected, the target-related data is processed based on the user tag rule engine to obtain at least one target word;

[0041] Determine the information dimension corresponding to each target word, where the information dimension includes any one of a risk information dimension, an asset information dimension, and a communication information dimension;

[0042] Determine the priority of each target word according to its information dimension;

[0043] At least one target word is sorted in order according to the priority, wherein the target word with a high priority is sorted first, and the target word with a low priority is sorted last.

[0044] In a second aspect, an embodiment of the present application provides a system for building a user tag rule engine, the system comprising:

[0045] The word segmentation module is used to perform word segmentation processing on the historical associated data of historical users to obtain multiple word segments;

[0046] A determination module, configured to determine a plurality of segmented word groups from the plurality of segmented words based on the similarities between the plurality of segmented words, to form a first word library including the plurality of segmented word groups; wherein each segmented word group includes at least two segmented words whose similarities satisfy a preset similarity condition;

[0047] A generating module, configured to generate, based on the first word library, a plurality of search rules corresponding to a plurality of word groups, wherein different word groups correspond to different search rules, and each search rule is used to determine a user tag of any word in the word group to which it corresponds;

[0048] The construction module is used to construct a user tag rule engine according to multiple search rules, and the user tag rule engine is used to determine the user tag of the user.

[0049] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and when the computer program is executed by the processor, the method for constructing a user tag rule engine provided in the first aspect is implemented.

[0050] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by one or more processors, the method for building a user tag rule engine provided in the first aspect is implemented.

[0051] The method for constructing a user tag rule engine provided by the embodiment of the present application can perform word segmentation processing on the historical associated data of historical users to obtain multiple word segments. Then, based on the similarity between the multiple word segments, multiple word segmentation groups are determined from the multiple word segments to form a first word library including multiple word segmentation groups, and multiple search rules corresponding to the multiple word segmentation groups are generated according to the first word library. The search rules corresponding to different word segmentation groups are different, and each search rule is used to determine the user tag of any word segmentation in the word segmentation group to which it corresponds. Finally, according to the multiple search rules, a user tag rule engine for determining user tags is constructed. The present application is based on multiple search rules generated by the first word library to construct a user tag rule engine, and all word segments of the same word segmentation group in the first word library correspond to the same search rule. In this way, when facing complex and diverse user data in banking business, such as word segments with high similarity, it can reduce the need to expand the target rules corresponding to each word segmentation by writing new code to adapt to the ever-changing business needs, thereby reducing the development time cost and labor cost in the process of determining user tags. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. It should be understood that the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 A flowchart of a method for building a user tag rule engine provided in an embodiment of the present application;

[0054] Figure 2 A word segmentation processing flow chart included in a method for building a user tag rule engine provided in an embodiment of the present application;

[0055] Figure 3 A flowchart of determining word groups included in a method for building a user tag rule engine provided in an embodiment of the present application;

[0056] Figure 4 A sorting flow chart involved in a method for building a user tag rule engine provided in an embodiment of the present application;

[0057] Figure 5 A schematic diagram of functional modules of a system for constructing a user tag rule engine provided in an embodiment of the present application;

[0058] Figure 6 A diagram of the internal structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0060] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0061] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0062] In various embodiments of the present application, the expression "or" or "at least one of A or / and B" includes any combination or all combinations of the words listed at the same time. For example, the expression "A or B" or "at least one of A or / and B" may include A, may include B, or may include both A and B.

[0063] In the description of the present application, it should be noted that if the terms "upper", "lower", "inside", "outside", etc. appear to indicate an orientation or position relationship, it is based on the orientation or position relationship shown in the drawings, or is the orientation or position relationship in which the product of the invention is usually placed when used. It is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0064] In addition, the terms “first”, “second”, etc., if used, are merely used to distinguish between the descriptions and should not be understood as indicating or implying relative importance.

[0065] It should be noted that, in the absence of conflict, the features in the embodiments of the present application may be combined with each other.

[0066] Furthermore, in the embodiments of the present application, the term "connection" may refer to "electrical connection" or "direct connection". "Electrical connection" may refer to direct electrical connection between two components or electrical connection between two components via one or more normally open tubes or other components.

[0067] To facilitate a better understanding of the solutions of the embodiments of the present application, the relevant technologies are first introduced below.

[0068] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0069] Natural language processing (NLP): NLP uses computers to process, understand and apply human languages ​​(such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary subject between computer science and linguistics. It is often referred to as computational linguistics. Natural language processing includes grammatical analysis, semantic analysis, and text understanding. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining. It involves data mining related to language processing, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computing.

[0070] Information Extraction: A text processing technology that extracts specified types of entity, relationship, event and other factual information from natural language text and forms structured data output. Information extraction is a technology that extracts specific information from text data. Text data is composed of some specific units, such as sentences, paragraphs, and chapters. Text information is composed of some small specific units, such as characters, words, phrases, sentences, paragraphs, or a combination of these specific units. Extracting noun phrases, names, place names, etc. from text data are all text information extraction. Of course, the information extracted by text information extraction technology can be various types of information.

[0071] Optical Character Recognition (OCR) technology refers to the process in which an electronic device (such as a scanner or digital camera) examines characters printed on paper, determines their shapes by detecting dark and light patterns, and then uses character recognition methods to translate the shapes into computer text; that is, for printed characters, an optical method is used to convert the text in a paper document into a black and white dot matrix image file, and recognition software is used to convert the text in the image into text format for further editing and processing by word processing software.

[0072] The rule engine evolved from the inference engine and is a component embedded in the application. It separates business decisions from application code and uses predefined semantic modules to write business decisions. It receives data input, interprets business rules, and makes business decisions based on business rules.

[0073] In order to solve the technical problems in the background technology, the embodiments of the present application provide a method, system, device and storage medium for constructing a user tag rule engine. The method for constructing a user tag rule engine provided by the embodiments of the present application is first introduced below.

[0074] See also Figure 1 , Figure 1 A flowchart of a method for constructing a user tag rule engine provided in an embodiment of the present application. The method for constructing a user tag rule engine can be applied to a system and an electronic device for constructing a user tag rule engine in the following embodiments, wherein the electronic device includes a personal computer, a server, a mobile device, a cloud computing platform, and a supercomputer, etc. The method for constructing a user tag rule engine will be introduced from the perspective of application to an electronic device. The method specifically includes the following steps:

[0075] Step 110, performing word segmentation processing on the historical associated data of the historical users to obtain a plurality of word segments.

[0076] Step 120: Based on the similarities between the multiple segmented words, multiple segmented word groups are determined from the multiple segmented words to form a first word library including the multiple segmented word groups, wherein each segmented word group includes at least two segmented words whose similarities meet a preset similarity condition.

[0077] Step 130: Generate multiple search rules corresponding to multiple word groups based on the first word library. Different word groups correspond to different search rules, and each search rule is used to determine the user tag of any word in the corresponding word group.

[0078] Step 140: construct a user tag rule engine according to the multiple search rules, and the user tag rule engine is used to determine the user tag of the user.

[0079] The embodiments of the present application can acquire and process relevant data such as historical association data based on artificial intelligence technology. Among them, artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0080] The basic AI technologies associated with AI generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0081] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0082] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0083] The embodiment of the present application provides a method for constructing a user tag rule engine, which can perform word segmentation processing on the historical associated data of historical users to obtain multiple word segments. Then, based on the similarity between the multiple word segments, multiple word segmentation groups are determined from the multiple word segments to form a first word library including multiple word segmentation groups, and multiple search rules corresponding to the multiple word segmentation groups are generated according to the first word library. The search rules corresponding to different word segmentation groups are different, and each search rule is used to determine the user tag of any word segmentation in the word segmentation group to which it corresponds. Finally, according to the multiple search rules, a user tag rule engine for determining user tags is constructed. The present application is based on multiple search rules generated by the first word library to construct a user tag rule engine, and all word segments of the same word segmentation group in the first word library correspond to the same search rule. In this way, when facing complex and diverse user data in banking business, such as word segments with high similarity, it can reduce the need to expand the target rules corresponding to each word segmentation by writing new code to adapt to the ever-changing business needs, thereby reducing the development time cost and labor cost in the process of determining user tags.

[0084] The following will be Figure 1 Each step of the method is described in detail.

[0085] In step 110, the electronic device may perform word segmentation processing on the historical association data to obtain a plurality of word segments.

[0086] The historical user can be a user within the past year or within the past five years, and can be selected according to actual needs, which is not limited in this embodiment. The historical associated data is data corresponding to the historical user. Taking a bank user as an example, the historical associated data includes the input information of the bank user when handling business in the bank, the visit information of the bank business personnel when visiting the user outside, etc.

[0087] Historically relevant data can be stored on paper, in images, or in electronic files.

[0088] Word segmentation is a basic step in NLP, which involves breaking down text strings into smaller units, which can be words, phrases, or other meaningful symbols. The purpose of word segmentation is to enable electronic devices to understand the structure of human language for further processing and analysis. Electronic devices can determine how to obtain historical association data based on the storage form of historical association data, and then perform word segmentation on the historical association data after obtaining it.

[0089] For example, if the historical association data is stored in the form of paper, the electronic device may use optical character recognition (OCR) technology to identify the historical association data on the paper, and extract information from the historical association data in combination with NLP technology, and then perform word segmentation processing after obtaining the historical association data.

[0090] In some embodiments, the number of user types of historical users and the number of users corresponding to each user type can be used to determine whether to select users within the past year or within the past five years. Exemplarily, if the number of user types in the historical users within the past five years is less than 30, it can be said that the user types of the historical users in the past five years are relatively single. Accordingly, the probability of duplication of historical associated data within the past five years is high, so the electronic device can select users with a relatively close time, such as within the past year, as historical users in this implementation, which can improve the stability of building a user tag rule engine and the accuracy of the user tag rule engine in determining user tags.

[0091] In some embodiments, word segmentation processing can be achieved through machine learning algorithms such as Hidden Markov Model (HMM), Conditional Random Field (CRF), etc.

[0092] In some embodiments, word segmentation processing can be implemented through a recurrent neural network (RNN).

[0093] In step 120, the similarity can represent the degree of approximation between at least two word segments. The electronic device can divide all word segments according to the similarity between different word segments, so as to divide the word segments with high similarity into the same word segment group, and multiple word segment groups form a first word library.

[0094] The preset similarity conditions include:

[0095] The similarity is greater than or equal to a preset similarity threshold, wherein multiple preset similarity thresholds can be set.

[0096] The similarity threshold can be set according to actual needs and is not limited in this embodiment.

[0097] In some embodiments, the electronic device can determine the cosine similarity between at least two segmented words based on the word vectors of each segmented word, and determine multiple segmented word groups from multiple segmented words based on the cosine similarity between at least two segmented words. Specifically, cosine similarity is also called cosine similarity, which evaluates the similarity of two vectors by calculating the cosine value of the angle between them.

[0098] In some embodiments, the electronic device may determine the similarity between at least two word segments through a group of related models used to generate word vectors, such as through a Word2vec model.

[0099] In step 130, the search rules may be used to determine the user tags corresponding to the segmented words, and each segmented word group may correspond to a search rule.

[0100] User tags can be understood as data points used to describe user characteristics and behaviors, which can help companies such as banks better understand their customers and provide customers with more personalized services and products accordingly.

[0101] In some embodiments, user tags can be divided into consumption data tags, behavior preference tags, basic attribute tags, user asset tags, etc. by tag type. Among them, consumption data tags may include consumption frequency, user value, etc., behavior preference tags may include consumption day preference, consumption day time period preference, etc., basic attribute tags may include user gender, user membership level, etc., and user asset tags may include user member available balance, user member available points, user member available coupons, etc.

[0102] In some embodiments, generating multiple search rules corresponding to multiple word groups includes randomly selecting a word from each word group, and then generating a search rule corresponding to the word group where the word belongs based on the randomly selected word.

[0103] Exemplarily, for a word group A (a1, a2, ..., a10), where a1 to a10 are 10 word groups in word group A, a random algorithm can be used to randomly select one word, such as a8, from these 10 word groups, and then determine the corresponding search rule based on word group a8. Specifically, if word group a8 is "reject", the result corresponding to word group a8 is "invalid customer", that is, it can be understood that word group a8 and the result corresponding to word group a8 form a search rule, and the user label of word group A determined by the search rule is "invalid customer". There is no need to determine the corresponding search rules for other word groups in subgroup A, which can reduce the need to expand the search rules corresponding to each word group by writing new code.

[0104] In some embodiments, multiple search rules corresponding to multiple word groups are generated, including selecting a word with high importance from each word group. Specifically, a graph model of each word group can be constructed using the TextRank algorithm, and nodes with high importance in the graph model can be identified through voting or recommendation mechanisms, and each node of the graph model can correspond to a word.

[0105] In step 140, after a plurality of search rules have been determined through the above embodiments, the electronic device may further construct a user tag rule engine based on these search rules, through which the user tag rule engine may be used to quickly and reliably determine the user tag.

[0106] In some embodiments, the electronic device can add multiple search rules to a preset rule engine matcher, and then implement the matching of the search rules through the preset rule engine matcher. The preset rule engine matcher can be configured with a regular expression to implement the matching of the search rules.

[0107] In some embodiments, the electronic device may pre-process the historical associated data of the historical user. The pre-processing includes data validity and integrity checks, and if one of the checks fails, directly using a preset user tag as the user tag of the historical user, such as "invalid user" or "low-value user".

[0108] See also Figure 2 , Figure 2 A word segmentation processing flow chart included in a method for building a user tag rule engine provided in an embodiment of the present application.

[0109] In a possible implementation, step 110 performs word segmentation processing on the historical associated data of the historical user, and obtaining multiple word segments may include but is not limited to steps 1110 to 1130:

[0110] Step 1110, obtaining user portrait data of historical users.

[0111] Step 1120, forming target word segmentation rules for historical users based on user portrait data.

[0112] Step 1130 , performing word segmentation processing on the historical associated data of the historical user according to the target word segmentation rule of the historical user to obtain a plurality of word segments.

[0113] This embodiment realizes the word segmentation processing of historical related data through the target word segmentation rules formed by the user portrait data, which can improve the accuracy of the word segmentation processing and further improve the accuracy and reliability of the subsequent construction of the user tag rule engine.

[0114] Specifically, user portrait data can also be regarded as a user model. For bank users, user portrait data may include the bank users' basic information, transaction behavior, asset status, credit record, risk preference, etc.

[0115] According to the user portrait data, the target word segmentation rules for historical users are formed, which can realize accurate word segmentation processing of historical user's historical related data. Specifically, if the user portrait data indicates that the corresponding user is a bank user, the corresponding target word segmentation rules can be executed based on the data in the financial field without the need to be processed according to the conventional word segmentation method.

[0116] For example, there are some specific words in the financial field, such as Loan Prime Rate (LPR), hedge funds, financial derivatives, etc. If the "loan market quotation rate" is segmented according to the general segmentation method, "loan", "duration", "quotation" and "interest rate" will generally be obtained. However, if it is determined that the user portrait data represents the identity of the user as a bank user, the target segmentation rule can be formed according to the relevant data in the financial field. When the "loan market quotation rate" is segmented according to the target segmentation rule, the "loan market quotation rate" will be obtained. The accuracy and efficiency of the segmentation process can be improved compared with the conventional segmentation method.

[0117] In some embodiments, the user portrait data may be information reserved by the electronic device user at a bank, or may be obtained by analyzing the user's account registration information.

[0118] In some embodiments, target word segmentation rules for historical users formed based on user portrait data can be integrated into the word segmenter, and the word segmenter can adopt the machine learning algorithm in the above embodiments.

[0119] In a possible implementation, before generating a plurality of search rules corresponding to a plurality of word groups according to the first word library, the method further includes:

[0120] Based on the word frequency of each segmented word, determine multiple target segmented words from the multiple segmented words to form a second word library including the multiple target segmented words;

[0121] According to the first word library, a plurality of search rules corresponding to the plurality of word groups are generated, including:

[0122] Generate a plurality of first search rules corresponding to a plurality of word groups according to the first word library;

[0123] Generating a plurality of second search rules corresponding to a plurality of target word segments according to the second word library;

[0124] The multiple search rules include multiple first search rules and multiple second search rules.

[0125] The embodiment of the present application introduces a second vocabulary library formed according to word frequency division, and jointly determines multiple search rules based on the first vocabulary library formed in the above embodiment and the second vocabulary library formed in this embodiment, thereby further improving the accuracy and reliability of determining the search rules, and thus also being able to improve the reliability of the user tag rule engine.

[0126] Specifically, the word frequency can be used to characterize the frequency of occurrence of a segmentation in the historical association data. Based on the word frequency of each segmentation, multiple target segmentations are determined from multiple segmentations, including determining the word frequency of each segmentation in the multiple segmentations, and taking the segmentation whose word frequency is greater than or equal to the preset word frequency as the target segmentation. The preset word frequency can be set according to actual needs, and this embodiment does not limit it here.

[0127] In some embodiments, after generating a plurality of second search rules corresponding to a plurality of target word segments according to the second word library, the method further includes:

[0128] Repeatedly test the first search rule and the second search rule to obtain a test result;

[0129] When the detection result indicates that the same search rule exists, a duplicate search rule is selected from the first search rule or the second search rule and deleted.

[0130] In a possible implementation, a plurality of first search rules corresponding to a plurality of word groups are generated according to the first word library, including:

[0131] The first word library is analyzed by using a preset deep learning model to obtain feature data of each word segment in the first word library, wherein the feature data of each word segment includes semantic information and contextual relationship information of each word segment;

[0132] Generate a plurality of first search rules according to the semantic information and contextual relationship information of each word segment;

[0133] According to the second word library, a plurality of second search rules corresponding to the plurality of second target word segments are generated, including:

[0134] The second word library is analyzed by using a preset deep learning model to obtain feature data of each target word segment in the second word library, wherein the feature data of each target word segment includes semantic information and contextual relationship information of each word segment;

[0135] Generate multiple second search rules according to the semantic information and contextual relationship information of each target word segment;

[0136] Among them, the preset deep learning models include a recurrent neural network model, a transformer model, a natural language processing model and a recurrent neural network model.

[0137] This embodiment obtains feature data including semantic information and contextual relationship information of each word segment by presetting a deep learning neural network, and generates a first search rule corresponding to the first vocabulary and a second search rule corresponding to the second vocabulary based on the feature data, which can further improve the accuracy and reliability of determining the first search rule and the second search rule.

[0138] Specifically, semantic information can be used to indicate the meaning conveyed by a segmentation. Contextual relationship information can be used to determine whether a segmentation constitutes an independent word, or to characterize the specific meaning of a segmentation in a specific context.

[0139] Using a preset deep learning model to determine the semantic information and contextual relationship information of word segmentation can more accurately handle ambiguous problems that are prone to occur in word segmentation.

[0140] The preset deep learning model may also include other models besides the models listed in this embodiment, which will not be introduced in detail in this embodiment.

[0141] In a possible implementation, based on the word frequency of each segmented word, multiple target segmented words are determined from multiple segmented words to form a second word library including the multiple target segmented words, including:

[0142] According to the frequency of each participle, the weight value of each participle is determined, wherein the frequency of the word is directly proportional to the weight value;

[0143] When the source channels of each word segment are determined, the word segments with the same source channel are assigned to the same data group to form an initial data set, wherein the initial data set includes at least one data group;

[0144] Get the co-occurrence frequency between different word segments in each data group;

[0145] Determine a first sub-participle in each data group, where the first sub-participle is a participle corresponding to a co-occurrence frequency greater than or equal to a preset frequency;

[0146] The first sub-participle whose weight value is greater than or equal to the preset weight value in each data group is used as the second sub-participle;

[0147] All second sub-words in all data groups are used as multiple target word segments to form a second word library including the multiple target word segments.

[0148] This embodiment determines a second vocabulary including multiple target word segments through the weight values ​​of each word segment and the co-occurrence frequency between different word segmentations. The influence relationship of the first word segment on the second word segment is formed through the weight values ​​and the co-occurrence frequency, which can further improve the accuracy and reliability of determining the second word segment including multiple target word segmentations.

[0149] Specifically, the word frequency is in direct proportion to the weight value. It can be understood that the greater the word frequency of a segmentation is, the greater the weight value of the segmentation is.

[0150] The co-occurrence frequency can be used to indicate the number of times or probabilities that different sub-words co-occur in the same context. In this embodiment, the first sub-word can indicate that the number of times or frequency that co-occurs in the same context is greater than or equal to the preset frequency. Similarly, the probability that the first sub-word is a synonym or a near synonym is relatively high.

[0151] In some embodiments, source channels include online channels and offline channels.

[0152] The preset frequency and the preset weight value can be set according to actual needs, and this embodiment does not limit them here.

[0153] See also Figure 3 , Figure 3 A flowchart of determining word groups included in a method for building a user tag rule engine provided in an embodiment of the present application.

[0154] In a possible implementation, step 120 determines multiple word groups from multiple word segments based on the similarities between the multiple word segments, including but not limited to steps 1210 to 1220:

[0155] Step 1210: Determine a plurality of preset similarity threshold ranges.

[0156] Step 1220: Based on the similarities between the multiple segmented words, the segmented words corresponding to the similarities within each preset similarity threshold range are respectively formed into a segmented word group to obtain multiple segmented word groups.

[0157] This embodiment determines multiple word groups by using multiple preset similarity threshold ranges and the similarities between multiple similarities, which can improve the efficiency of obtaining multiple word groups. The specific implementation process is the same or similar to the implementation process of the preset similarity threshold in the above embodiment, and will not be repeated here.

[0158] See also Figure 4 , Figure 4 A sorting flowchart involved in a method for building a user tag rule engine provided in an embodiment of the present application.

[0159] In a possible implementation, after constructing a user tag rule engine according to multiple search rules, the method further includes steps 210 to 240:

[0160] Step 210 , when target association data of a newly added user is detected, the target association data is processed based on a user tag rule engine to obtain at least one target word.

[0161] Step 220, determining the information dimension corresponding to each target word, where the information dimension includes any one of a risk information dimension, an asset information dimension, and a communication information dimension.

[0162] Step 230: Determine the priority of each target word according to the information dimension of each target word.

[0163] Step 240 , sorting at least one target word according to the priority level, wherein the target word with a high priority is sorted first, and the target word with a low priority is sorted last.

[0164] This embodiment determines the priority of the target vocabulary and sorts all the target vocabulary according to the priority. The target vocabulary can be understood as the user tag of the new user determined by the user tag rule engine. After the target vocabulary is sorted, it is convenient for the relevant business personnel of the bank to quickly analyze the new users according to the sorted target vocabulary.

[0165] The target association data is similar to the historical association data in the above embodiment, except that the corresponding users are different. For the introduction of the target association data, reference can be made to the introduction of the historical association data in the above embodiment, which will not be repeated here.

[0166] The target association data is processed based on the user tag rule engine to obtain at least one target word. This process can be understood as the process of determining the user tag of the newly added user through the user tag rule engine. The at least one target word can be regarded as at least one user tag of the newly added user.

[0167] In some embodiments, the priorities corresponding to the risk information dimension, the asset information dimension, and the communication information dimension decrease in sequence.

[0168] In some embodiments, in the process of processing target associated data based on the user tag rule engine, if there are new word segments in the target associated data that are not stored in the first vocabulary and the second vocabulary, the new word segments can be analyzed according to the operations in the aforementioned embodiments and stored in the first vocabulary or the second vocabulary to achieve an update of the first vocabulary or the second vocabulary.

[0169] Specifically, if the new word segmentation is stored in a word segmentation group that already exists in the first word library, it can be indicated that the new word segmentation has a high similarity with a word segmentation group in the first word library, and there is no need to generate a new search rule. If the new word segmentation is stored in a new word segmentation group in the first word library, a search rule corresponding to the new word segmentation group is generated, and the user tag rule engine can be rebuilt to achieve the expansion and update of the user tag rule engine.

[0170] Corresponding to the above method embodiment, the present application embodiment also provides a system for building a user tag rule engine, see Figure 5 , Figure 5A functional module diagram of a system for constructing a user tag rule engine provided in an embodiment of the present application, wherein the system 300 for constructing a user tag rule engine includes:

[0171] A word segmentation module 310 is used to perform word segmentation processing on the historical associated data of the historical user to obtain multiple word segments;

[0172] A determination module 320 is used to determine a plurality of segmentation groups from the plurality of segmentations based on the similarities between the plurality of segmentations, so as to form a first word library including the plurality of segmentation groups; wherein each segmentation group includes at least two segmentations whose similarities satisfy a preset similarity condition;

[0173] A generating module 330 is used to generate a plurality of search rules corresponding to a plurality of word groups according to the first word library, wherein different word groups correspond to different search rules, and each search rule is used to determine a user tag of any word in the word group corresponding to it;

[0174] The construction module 340 is used to construct a user tag rule engine according to multiple search rules, and the user tag rule engine is used to determine the user tag of the user.

[0175] The system for constructing a user tag rule engine provided in the embodiment of the present application can achieve the following Figure 1 The various processes implemented in the method embodiments can achieve similar or identical technical effects, and will not be described again here to avoid repetition.

[0176] In a possible implementation, the word segmentation module 310 is specifically used for:

[0177] Obtain user portrait data of historical users;

[0178] Based on user portrait data, target word segmentation rules for historical users are formed;

[0179] The historical associated data of the historical user is segmented according to the target segmentation rule of the historical user to obtain multiple segmentations.

[0180] In a possible implementation, the system 300 for constructing a user tag rule engine further includes a determination submodule, and the determination submodule is specifically used for:

[0181] Based on the word frequency of each segmented word, determine multiple target segmented words from the multiple segmented words to form a second word library including the multiple target segmented words;

[0182] The generation module 330 is further specifically used for:

[0183] Generate a plurality of first search rules corresponding to a plurality of word groups according to the first word library;

[0184] Generating a plurality of second search rules corresponding to a plurality of target word segments according to the second word library;

[0185] The multiple search rules include multiple first search rules and multiple second search rules.

[0186] In a possible implementation, the generation module 330 further includes a first generation submodule and a second generation submodule, and the first generation submodule is specifically used to:

[0187] The first word library is analyzed by using a preset deep learning model to obtain feature data of each word segment in the first word library, wherein the feature data of each word segment includes semantic information and contextual relationship information of each word segment;

[0188] Generate a plurality of first search rules according to the semantic information and contextual relationship information of each word segment;

[0189] The second generation submodule is specifically used for:

[0190] The second word library is analyzed by using a preset deep learning model to obtain feature data of each target word segment in the second word library, wherein the feature data of each target word segment includes semantic information and contextual relationship information of each word segment;

[0191] Generate multiple second search rules according to the semantic information and contextual relationship information of each target word segment;

[0192] Among them, the preset deep learning models include a recurrent neural network model, a transformer model, a natural language processing model and a recurrent neural network model.

[0193] In a possible implementation, the word segmentation module 310 includes a word segmentation submodule, which is specifically used for:

[0194] According to the frequency of each participle, the weight value of each participle is determined, wherein the frequency of the word is directly proportional to the weight value;

[0195] When the source channels of each word segment are determined, the word segments with the same source channel are assigned to the same data group to form an initial data set, wherein the initial data set includes at least one data group;

[0196] Get the co-occurrence frequency between different word segments in each data group;

[0197] Determine a first sub-participle in each data group, where the first sub-participle is a participle corresponding to a co-occurrence frequency greater than or equal to a preset frequency;

[0198] The first sub-participle in each data group that is greater than or equal to a preset weight value is used as the second sub-participle;

[0199] All second sub-words in all data groups are used as multiple target word segments to form a second word library including the multiple target word segments.

[0200] In a possible implementation manner, the determination module 320 is further specifically configured to:

[0201] determining a plurality of preset similarity threshold ranges;

[0202] Based on the similarity between the multiple segmented words, the segmented words corresponding to the similarity within each preset similarity threshold range are respectively formed into a segmented word group to obtain multiple segmented word groups.

[0203] In a possible implementation, the system 300 for constructing a user tag rule engine further includes a processing module, and the processing module is specifically used to:

[0204] When target-related data of a newly added user is detected, the target-related data is processed based on the user tag rule engine to obtain at least one target word;

[0205] Determine the information dimension corresponding to each target word, where the information dimension includes any one of a risk information dimension, an asset information dimension, and a communication information dimension;

[0206] Determine the priority of each target word according to its information dimension;

[0207] At least one target word is sorted in order according to the priority, wherein the target word with a high priority is sorted first, and the target word with a low priority is sorted last.

[0208] The present application also provides an electronic device. The present application also provides an electronic device. Figure 6 , Figure 6 An internal structure diagram of an electronic device provided for an embodiment of the present application. The electronic device includes a processor, a memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the electronic device stores an operating system and may also store a computer program, which, when executed by the processor, enables the processor to implement the method for building a user tag rule engine applied to the electronic device in the above-mentioned embodiment. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to execute the method for building a user tag rule engine. Those skilled in the art will appreciate that Figure 6 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0209] The embodiment of the present application also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for building a user tag rule engine in the method embodiment is implemented.

[0210] An embodiment of the present application provides a computer program product, which is stored in a storage medium. The program product is executed by at least one processor to implement the various processes of the embodiment of the method for constructing a user tag rule engine as described above, and can achieve similar or identical technical effects. To avoid repetition, it will not be described here.

[0211] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0212] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for constructing a user tag rule engine, characterized in that: The method comprises: Perform word segmentation on the historical associated data of historical users to obtain multiple word segments; Based on the similarities between the multiple participles, determine multiple participle groups from the multiple participles to form a first word library including the multiple participle groups, wherein each of the participle groups includes at least two participles whose similarities meet a preset similarity condition; Generate a plurality of search rules corresponding to the plurality of word groups according to the first word library, wherein different word groups correspond to different search rules, and each search rule is used to determine a user tag of any word in the word group corresponding to it; A user tag rule engine is constructed according to the multiple search rules, and the user tag rule engine is used to determine the user tag of the user.

2. The method according to claim 1, characterized in that The historical user's historical associated data is segmented to obtain multiple segmented words, including: Obtaining user portrait data of the historical user; According to the user portrait data, forming a target word segmentation rule for the historical user; The historical associated data of the historical user is segmented according to the target segmentation rule of the historical user to obtain the multiple segmentations.

3. The method according to claim 1, characterized in that Before generating a plurality of search rules corresponding to the plurality of word groups according to the first word library, the method further comprises: Based on the word frequency of each of the segmented words, determine a plurality of target segmented words from the plurality of segmented words, and form a second word library including the plurality of target segmented words; The step of generating a plurality of search rules corresponding to the plurality of word groups according to the first word library includes: Generating a plurality of first search rules corresponding to the plurality of word groups according to the first word library; Generating a plurality of second search rules corresponding to the plurality of target word segments according to the second word library; The multiple search rules include the multiple first search rules and the multiple second search rules.

4. The method according to claim 3, characterized in that Generating a plurality of first search rules corresponding to the plurality of word groups according to the first word library includes: Using a preset deep learning model to analyze the first word library, obtain feature data of each word segment in the first word library, wherein the feature data of each word segment includes semantic information and contextual relationship information of each word segment; Generating the plurality of first search rules according to the semantic information and contextual relationship information of each of the segmented words; Generating a plurality of second search rules corresponding to the plurality of second target word segmentations according to the second word library includes: The second word library is analyzed by using a preset deep learning model to obtain feature data of each target word in the second word library, wherein the feature data of each target word includes semantic information and contextual relationship information of each word; Generating the plurality of second search rules according to the semantic information and contextual relationship information of each of the target word segments; Among them, the preset deep learning model includes a recurrent neural network model, a transformer model, a natural language processing model and a recurrent neural network model.

5. The method according to claim 3, characterized in that: The step of determining a plurality of target segmented words from the plurality of segmented words based on the word frequencies of the respective segmented words, and forming a second word library including the plurality of target segmented words, comprises: Determine the weight value of each of the participles according to the word frequency of each of the participles, wherein the word frequency is in direct proportion to the weight value; When the source channels of each of the segmented words are determined, the segmented words with the same source channel are assigned to the same data group to form an initial data set, wherein the initial data set includes at least one of the data groups; Obtaining the co-occurrence frequency between different word segments in each of the data groups; Determine a first sub-participle in each of the data groups, the first sub-participle being a participle corresponding to a co-occurrence frequency greater than or equal to a preset frequency; Taking the first sub-participle in each of the data groups whose weight value is greater than or equal to the preset weight value as the second sub-participle; All of the second sub-words in all of the data groups are used as the multiple target word segments to form a second word library including the multiple target word segments.

6. The method according to claim 1, characterized in that The determining a plurality of word groups from the plurality of word segments based on the similarities between the plurality of word segments comprises: determining a plurality of preset similarity threshold ranges; Based on the similarities between the multiple segmented words, the segmented words corresponding to the similarities within each of the preset similarity thresholds are respectively formed into a segmented word group to obtain the multiple segmented word groups.

7. The method according to claim 1, characterized in that After constructing the user tag rule engine according to the multiple search rules, the method further includes: When target-related data of a newly added user is detected, the target-related data is processed based on the user tag rule engine to obtain at least one target word; Determine the information dimension corresponding to each of the target words, the information dimension including any one of a risk information dimension, an asset information dimension, and a communication information dimension; Determining the priority of each target word according to the information dimension of each target word; The at least one target word is sorted in order according to the priority, wherein the target word with a high priority is sorted first, and the target word with a low priority is sorted last.

8. A system for building a user tag rule engine, characterized in that: The system comprises: The word segmentation module is used to perform word segmentation processing on the historical associated data of historical users to obtain multiple word segments; A determination module, configured to determine a plurality of segmentation groups from the plurality of segmentations based on the similarities between the plurality of segmentations, and form a first word library including the plurality of segmentation groups; wherein each of the segmentation groups includes at least two segmentations whose similarities satisfy a preset similarity condition; A generating module, configured to generate, based on the first word library, a plurality of search rules corresponding to the plurality of word groups, wherein different word groups correspond to different search rules, and each search rule is used to determine a user tag of any word in the word group to which it corresponds; A construction module is used to construct a user tag rule engine according to the multiple search rules, and the user tag rule engine is used to determine the user tag of the user.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein a computer program is stored in the memory, and the computer program implements the method according to any one of claims 1 to 7 when executed by the processor.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by one or more processors, the method according to any one of claims 1 to 7 is implemented.

Images