CN116681561A

CN116681561A - Policy matching method and device, electronic equipment and storage medium

Info

Publication number: CN116681561A
Application number: CN202310451440.8A
Authority: CN
Inventors: 程云辉; 高晓丽; 姚伟华; 蒋申为; 吴晓晴
Original assignee: Shanghai E&p International Inc
Current assignee: Shanghai E&p International Inc
Priority date: 2023-04-24
Filing date: 2023-04-24
Publication date: 2023-09-01

Abstract

The embodiment of the application discloses a policy matching method, a policy matching device, electronic equipment and a storage medium. In the embodiment of the application, the policy file and the enterprise reporting information of the user can be respectively extracted in a structuring way to obtain the policy key information and the enterprise key information of a plurality of different structure types; the policy key information of the first structure type is precisely matched with the enterprise key information of the first structure type, so that precise matching degree is obtained; fuzzy matching is carried out on the policy key information of the second structure type and the enterprise key information of the second structure type, so that fuzzy matching degree is obtained; and determining the policy matching degree of the enterprise declaration information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree. The application improves the resolution of the policy, helps enterprises to be related to more preferential policies, saves the cost of manually gathering information, and simultaneously ensures that the whole policy is matched to form effective closed-loop optimization.

Description

Policy matching method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of information processing technologies, and in particular, to a policy matching method, a policy matching device, an electronic device, and a storage medium.

Background

Policies are a series of industry and enterprise development supporting measures established by government according to the economic needs of social development in combination with practical situations. To promote economic development throughout the area, government authorities may issue policy documents with rewards in platform websites for business and personal reporting. The policy file generally contains specific rewarding information and corresponding condition information, wherein the rewarding information refers to the offers which can be actually obtained by enterprises, such as monetary subsidy, tax subsidy and the like; the condition information is the qualification conditions such as operation place, annual income amount and the like which the enterprise or the individual needs to meet when reporting. The main purpose and function of policy matching is to make the policy reach the enterprise quickly and make the enterprise know the policy passively; the enterprise magnitude range which can be reached before and after the release of the dynamic prediction policy of the auxiliary government is provided, and effective data sample support is provided for budget frame establishment of part of policies.

However, in the actual use process, the existing policy matching method cannot perfectly match all condition information in the file, ignores behavior description sentences existing in the policy conditions, and adopts a traditional simple character similar matching mode, and the optimization mode of the traditional simple character similar matching mode needs to update rules according to new samples manually, so that the problems of high labor cost, low updating efficiency and the like are solved.

Disclosure of Invention

The embodiment of the application provides a policy matching method, a device, electronic equipment and a storage medium, which can perfectly match all condition information in a policy file, improve the resolution of the policy, save labor cost and have high information updating efficiency.

In a first aspect, an embodiment of the present application provides a policy matching method, including: respectively carrying out structured extraction on the policy file and enterprise reporting information of the user to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types; the number of different structure types includes a first structure type and a second structure type; the policy key information of the first structure type is precisely matched with the enterprise key information of the first structure type, so that precise matching degree is obtained; fuzzy matching is carried out on the policy key information of the second structure type and the enterprise key information of the second structure type, so that fuzzy matching degree is obtained; and determining the policy matching degree of the enterprise declaration information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

In some embodiments, the structured extraction is performed on the policy file and the enterprise declaration information of the user, so as to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types, including: acquiring a policy file; performing label structured extraction on the policy file to obtain label type policy key information; and carrying out corpus structured extraction on the policy file to obtain corpus type policy key information.

In some embodiments, the structured extraction is performed on the policy file and the enterprise declaration information of the user, so as to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types, including: acquiring enterprise declaration information of a user; carrying out label structured extraction on enterprise declaration information of a user to obtain label type enterprise key information; and carrying out corpus structured extraction on enterprise declaration information of the user to obtain corpus enterprise key information.

In some embodiments, the tag class policy key information and the tag class business key information are tagged with: the form of the label value is stored in a database; the corpus policy key information and the corpus enterprise key information are classified into: the text is stored in a database.

In some embodiments, the policy key information of the first structure type is accurately matched with the enterprise key information of the first structure type to obtain an accurate matching degree, where the first structure type is a tag type, and the method includes: and carrying out accurate matching on the label type policy key information and the label type enterprise key information to obtain accurate matching degree.

In some embodiments, performing accurate matching on the tag policy key information and the tag enterprise key information to obtain an accurate matching degree, including: when the label type is the ith label type, acquiring a first label value corresponding to the label type enterprise key information of which the label type is the ith label type and a second label value corresponding to the label type policy key information of which the label type is the ith label type; when the ith threshold condition is met between the first label value and the second label value, the accurate matching between the label enterprise key information with the label type of the ith label type and the label policy key information with the label type of the ith label type is judged, wherein i is more than or equal to 1.

In some embodiments, fuzzy matching is performed on policy key information of a second structure type and enterprise key information of the second structure type to obtain fuzzy matching degree, wherein the second structure type is a corpus type, and the method includes: and carrying out fuzzy matching on the corpus policy key information and the corpus enterprise key information to obtain fuzzy matching degree.

In some embodiments, performing fuzzy matching on the corpus-class policy key information and the corpus-class enterprise key information to obtain fuzzy matching degree includes: inputting corpus-type policy key information into a first BERT semantic model of a text similarity double-tower model, and outputting a policy text representation vector by the first BERT semantic model; inputting the corpus enterprise key information into a second BERT semantic model of the text similarity double-tower model, and outputting an enterprise text representation vector by the second BERT semantic model; and calculating the similarity between the policy text representation vector and the enterprise text representation vector to obtain the fuzzy matching degree.

In some embodiments, the policy matching degree includes a total matching degree between the user's business declaration information and the policy document, and a matching degree between the user's business declaration information and each condition item in the policy document; after determining the policy matching degree of the enterprise declaration information and the policy file of the user according to the accurate matching degree and the fuzzy matching degree, the method further comprises the following steps: acquiring the total matching degree between enterprise declaration information of a user and different policy files; ordering the total matching degree between the enterprise declaration information of the user and the different policy files from top to bottom to obtain a total matching degree ordering set; displaying the recommended policy to the user according to the total matching degree sequencing set; and according to the matching degree of the enterprise declaration information of the user and each condition item in the policy file, displaying the matching condition of the enterprise declaration information of the user to each condition item in the policy file on the policy detail page.

In some embodiments, according to the matching degree of the enterprise declaration information of the user and each condition item in the policy file, the matching condition of the enterprise declaration information of the user to each condition item in the policy file is displayed to the user on the policy detail page, and the method further includes: when detecting that the user checks the matching condition of each condition item in the policy file on the policy detail page, and the user clicks the first button through the policy detail page, saving an approval log; the approval log is a log approved by the user on the matching condition of each condition item of the policy file, and comprises a label log with similar policy-end corpus and enterprise-end corpus; when the fact that the user clicks the second button through the policy detail page is detected, enterprise declaration information of the user is updated, and a non-approval log is stored, wherein the non-approval log is a log which is not approved by the user for the matching condition of each condition item of the policy file, and the non-approval log comprises a label log which is dissimilar to the enterprise corpus at the policy end.

In some embodiments, the policy matching method further comprises: the label logs similar to the enterprise-side corpora are used as training sets; and training the text similarity double-tower model by using the training set to obtain a trained text similarity double-tower model.

In some embodiments, the policy matching method further comprises: when the fact that the user browses the policy detail page is detected, acquiring an information gain ratio according to the policy key information and the enterprise key information; generating a user questionnaire according to an information gain ratio, wherein the information gain ratio represents the priority of each question presentation in the user questionnaire; and updating enterprise declaration information of the user through the user questionnaire.

In a second aspect, an embodiment of the present application provides a policy matching device, including: the information extraction module is used for respectively carrying out structured extraction on the policy file and enterprise declaration information of the user to obtain a plurality of policy key information with different structure types and enterprise key information with different structure types; the number of different structure types includes a first structure type and a second structure type; the accurate matching module is used for accurately matching the policy key information of the first structure type with the enterprise key information of the first structure type to obtain accurate matching degree; the fuzzy matching module is used for carrying out fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain fuzzy matching degree; and the matching degree determining module is used for determining the policy matching degree of the enterprise declaration information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

In a third aspect, an embodiment of the present application provides an electronic device, including a memory storing a plurality of instructions; the processor loads instructions from the memory to perform steps in any of the policy matching methods provided by embodiments of the present application.

In a fourth aspect, embodiments of the present application provide a computer readable storage medium storing a plurality of instructions adapted to be loaded by a processor to perform steps in any of the policy matching methods provided by the embodiments of the present application.

The embodiment of the application can respectively and structurally extract the policy file and the enterprise declaration information of the user to obtain a plurality of policy key information with different structure types and enterprise key information with different structure types; then, based on the same structure type, carrying out precise matching or fuzzy matching on the policy key information and the enterprise key information to obtain precise matching degree and fuzzy matching degree; and determining the policy matching degree of the enterprise reporting information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

The application realizes an intelligent policy matching method, and in the traditional method, key information of policy conditions is only marked by labels: based on the value form representation, various types of policy condition texts are contained more perfectly by constructing a label system, so that the resolution degree of the policy is improved, and enterprises are helped to be associated with more preferential policies.

In addition, the application respectively develops an accurate matching module and a corpus text fuzzy matching module to finish comparison calculation aiming at the label class and corpus class condition items contained in a label system.

In addition, the application considers the requirement of further perfecting the enterprise information, intelligently generates the user questionnaire based on the information gain ratio of the policy and the enterprise information in the matching link, collects the enterprise user information, solves the problem of insufficient information, and simultaneously saves the cost of manually collecting the information.

The application also obtains the matching condition of the enterprise user information and the policy condition items by designing a user behavior log collection mechanism, and supervises and trains the collected similar label logs and dissimilar label logs to optimize a text similarity model, so that the whole matching module forms effective closed-loop optimization.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic view of a scenario of a policy matching method according to an embodiment of the present application;

FIG. 2a is a flowchart of a policy matching method according to an embodiment of the present application;

FIG. 2b is a schematic diagram of a policy tag architecture provided by an embodiment of the present application;

FIG. 2c is a schematic diagram of a text similarity double-tower model according to an embodiment of the present application;

FIG. 2d is a schematic diagram of a policy details page provided by an embodiment of the present application;

FIG. 2e is a schematic diagram of yet another policy details page provided by an embodiment of the present application;

FIG. 2f is a schematic diagram of yet another policy details page provided by an embodiment of the present application;

FIG. 2g is a block diagram of text similarity double-tower model optimization iterations provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a policy matching device according to an embodiment of the present application;

fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to fall within the scope of the application.

The embodiment of the application provides a policy matching method, a policy matching device, electronic equipment and a storage medium.

The policy matching device may be integrated in an electronic device, and the electronic device may be a terminal, a server, or other devices. The terminal can be a mobile phone, a tablet computer, an intelligent Bluetooth device, a notebook computer, a personal computer (Personal Computer, PC) or the like; the server may be a single server or a server cluster composed of a plurality of servers.

In some embodiments, the policy matching device may also be integrated in a plurality of electronic devices, for example, the policy matching device may be integrated in a plurality of servers, and the policy matching method of the present application is implemented by the plurality of servers.

In some embodiments, the server may also be implemented in the form of a terminal.

For example, referring to fig. 1, the electronic device may include a server 10, a storage terminal 11, a display 12, etc., where the storage terminal 11 may store a policy file and user's business reporting information, etc., the display may display a policy details page to the user, and the policy details page may display to the user a match of the user's business reporting information to each condition item in the policy file, etc., and the server 10, the storage terminal 11, and the display 12 are in communication connection with each other, which is not described herein.

Wherein the server 10 may include a processor, memory, and the like. The server 10 may firstly perform structured extraction on the policy file and the enterprise declaration information of the user to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types; the number of different structure types includes a first structure type and a second structure type; then, the policy key information of the first structure type is precisely matched with the enterprise key information of the first structure type, so that precise matching degree is obtained; performing fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain fuzzy matching degree; and finally, determining the policy matching degree of the enterprise reporting information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

The following will describe in detail. The numbers of the following examples are not intended to limit the preferred order of the examples.

In this embodiment, a policy matching method related to information processing is provided, as shown in fig. 2a, and the specific flow of the policy matching method may be as follows:

200. the server respectively carries out structured extraction on the policy file and enterprise declaration information of the user to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types.

In an embodiment of the present application, the number of different structure types includes a first structure type and a second structure type. The first structure type may be a tag type, the tag type may include a boolean class, a numerical class, a date class, a hierarchical class, and the like, and the second structure type may be a corpus type.

Specifically, the policy key information of the several different structure types may include tag-type policy key information, corpus-type policy key information, and the like, and the tag-type policy key information may include boolean-type policy key information, numerical-type policy key information, date-type policy key information, hierarchical-type policy key information, and the like. The enterprise-critical information of the several different structure types may include tag-type enterprise-critical information and corpus-type enterprise-critical information, and the tag-type enterprise-critical information may include boolean-type enterprise-critical information, numerical-type enterprise-critical information, date-type enterprise-critical information, hierarchical-type enterprise-critical information, and the like.

In the embodiment of the application, the structured extraction refers to extracting key element data capable of being structured from a file. Specifically, the label structuring extraction refers to extracting key element data capable of label structuring from a file, and the corpus structuring extraction refers to extracting key element data capable of corpus structuring from the file.

The embodiment of the application can firstly acquire the policy file, then perform label structured extraction on the policy file to obtain structured policy condition item label data, namely label type policy key information; the corpus structure extraction can be carried out on the policy file to obtain structured corpus data of the policy condition items, namely corpus type policy key information.

In the embodiment of the application, besides the structured extraction of the policy file, the structured extraction of the enterprise declaration information of the user can be performed. The method comprises the steps that a policy file can be obtained firstly, then label structuralization extraction is carried out on the policy file, and structured policy condition item label data, namely label type policy key information, is obtained; the corpus structure extraction can be carried out on the policy file to obtain structured corpus data of the policy condition items, namely corpus type policy key information.

In the embodiment of the application, in order to extract the key information in the policy file more perfectly and consider the condition text statement which exists in the policy and cannot be labeled, a policy label system is designed, wherein the policy label system can comprise labels in a value form text type and a label text type in a traditional simple matching calculation mode. As shown in fig. 2b, the structure types include boolean class, date class, numerical class, hierarchical class, corpus class, and the like. Specifically, the boolean type specification may be: whether the content of the policy condition item is satisfied is yes or no; examples of boolean types may be: "whether serious illegal loss of trust: and no (2). The date type description may be: the policy condition item is a date limit; date type samples may be: "date of business registration: after 1 month and 1 day 2022. The numerical type specification may be: the content of the policy condition item is an amount limit; examples of numerical types are: "real registered capital (ten thousand yuan): more than 5 hundred million yuan. The hierarchical type specification is: the content of the policy condition item comprises a plurality of text options, and the enterprise can meet the requirement of the text options; examples of layering types are: "business address: new patch area of temporary harbor "; the corpus types are described as follows: the above various types of policy condition item contents cannot be formed; corpus type samples may be: "Enterprise Activity: the international organization of introduction holds activities such as forum, peak meeting, etc. in the new area of the adjacent harbor. In other embodiments of the present application, an enterprise tag system is also designed, which is the same as or similar to the policy tag system, and will not be described in detail herein.

In the embodiment of the application, in the categories of the label system, boolean categories, date categories, numerical categories and layering categories are collectively called label categories, and the policy text sentences in the policy file, which cannot form the label categories, are corpus categories. The former can calculate matching based on rule judgment and character similarity comparison, and the latter realizes semantic understanding matching through a text similarity model.

210. And the server carries out accurate matching on the policy key information of the first structure type and the enterprise key information of the first structure type to obtain accurate matching degree.

In the embodiment of the present application, the first tag value may be a corresponding tag value of the label type i label type enterprise key information. The second tag value may be a tag value corresponding to tag class policy key information of which tag type is an i-th tag type.

Because policy and enterprise information both comprise label and corpus, the label information is labeled: tag value form and corpus information are classified into: the text is stored in the database, so that different modes are needed for matching the two types of information, the former is the accurate matching of the labels, and the latter is the fuzzy matching of the corpus.

Regarding the accurate matching of tag class information, the tag type can be specified to contain four types of Boolean class, numerical class, date class and layering class under the design framework of a tag system, and the matching modes are as follows:

boolean, comparing whether the tag values are consistent;

the value class compares whether the label value of the enterprise terminal is within the label value limit range under the policy condition;

the date class is used for comparing whether the label date value of the enterprise terminal is within the label date limit range under the policy condition;

and (3) layering, namely comparing whether the label value of the enterprise terminal is in a label value set under the policy condition, namely finishing the accurate matching of the label information through simple numerical comparison and character comparison.

Specifically, when the tag type is boolean and the tag value corresponding to the tag type boolean tag type enterprise key information is consistent with the tag value corresponding to the tag type boolean tag type policy key information, the tag type boolean tag type enterprise key information is precisely matched with the tag type boolean tag type policy key information.

When the label type is a numerical value type and the label value corresponding to the label type enterprise key information of the label type is in the label value limit range corresponding to the label type policy key information of the label type is a numerical value type, the label type enterprise key information of the label type is a numerical value type and the label type policy key information of the label type is a numerical value type are accurately matched.

When the label type is a date type and the label date value corresponding to the label type business key information of the date type is within the label date limit range corresponding to the label type policy key information of the date type, the label type business key information of the date type and the label type policy key information of the date type are accurately matched.

When the label type is a hierarchical type and the label value corresponding to the label type enterprise key information of the hierarchical type is in the label value set corresponding to the label type policy key information of the hierarchical type, the label type enterprise key information of the hierarchical type and the label type policy key information of the hierarchical type are accurately matched.

220. And the server performs fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain fuzzy matching degree.

In some embodiments, performing fuzzy matching on the corpus-class policy key information and the corpus-class enterprise key information to obtain fuzzy matching degree includes: inputting corpus-type policy key information into a first BERT (Bidirectional Encoder Representation from Transformers) semantic model of a text similarity double-tower model, and outputting a policy text representation vector by the first BERT semantic model; inputting the corpus enterprise key information into a second BERT semantic model of the text similarity double-tower model, and outputting an enterprise text representation vector by the second BERT semantic model; and calculating the similarity between the policy text representation vector and the enterprise text representation vector to obtain the fuzzy matching degree.

As shown in fig. 2c, in the embodiment of the present application, the text similarity double-tower model includes two BERT semantic models, a first BERT semantic model and a second BERT semantic model, respectively. The BERT semantic model is a pre-trained language representation model, and the bi-directional transformations are pre-trained using MLM (masked language model) to generate deep bi-directional language representations.

According to the embodiment of the application, for corpus information, matching calculation cannot be completed through simple numerical value and character comparison as in tag information, and semantic understanding comparison is needed by means of a natural language processing algorithm model. Therefore, the embodiment of the application acquires vector characterization of the whole sentence by constructing a text similarity double-tower model and transmitting an input text into a bottom BERT semantic model, for example, a policy end input text (namely, corresponding to corpus type policy key information) is input into a first BERT semantic model, and the first BERT semantic model outputs a policy text representation vector; the enterprise terminal inputs text (namely, corresponding to the corpus enterprise key information) into a second BERT semantic model, and the second BERT semantic model outputs an enterprise text representation vector; then, the similarity of the Cosine of the two sentence vectors is calculated as the fuzzy matching degree. The underlying BERT semantic model (i.e., the first BERT semantic model or the second BERT semantic model) is trained by a large amount of policy domain texts or enterprise declaration information, so as to improve the vector meaning representation of the policy domain vocabulary and improve the vector meaning representation of the enterprise declaration domain vocabulary.

In the text similarity double-tower model structure, the policy text expression vector and the enterprise text expression vector can be temporarily stored in a mode of calculation in advance, so that additional expenditure caused by repeated calculation is avoided, and an reasoning link can be realized by only obtaining the vector according to the input text. In addition, if a new text appears in the policy file, the new text has no corresponding text expression vector in temporary storage, the text expression vector corresponding to the new text is obtained only by executing the first BERT semantic model in the text similarity double-tower model, and then cosine similarity is calculated.

230. And the server determines the policy matching degree of the enterprise reporting information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

In the embodiment of the application, the matching degree of the enterprise and the policy can be obtained through intelligent matching of the policy, wherein the matching degree comprises the total matching degree of the enterprise and the policy and the matching degree of each condition item in the policy. The total matching degree is used for sequencing the displayed policies, and the recommended policies are displayed to enterprise users from top to bottom according to the matching degree. And the matching degree of each condition item in the policy is used for displaying the satisfaction condition of the specific policy detail page to each condition item to the enterprise user and simultaneously carrying out feedback collection.

In some embodiments, according to the matching degree of the enterprise declaration information of the user and each condition item in the policy file, the matching condition of the enterprise declaration information of the user to each condition item in the policy file is displayed to the user on the policy detail page, and the method further includes: when detecting that the user checks the matching condition of each condition item in the policy file on the policy detail page, and the user clicks the first button through the policy detail page, saving an approval log; the approval log comprises a label log similar to the enterprise-side corpus in terms of the policy-side corpus; when the fact that the user clicks the second button through the policy detail page is detected, enterprise declaration information of the user is updated, and a non-approval log is stored, wherein the non-approval log comprises label logs of which the policy-end corpus is dissimilar to the enterprise-end corpus.

In the embodiment of the application, the approval log is a log approved by the user for the matching condition of each condition item of the policy file, and the disapproval log is a log disapproval by the user for the matching condition of each condition item of the policy file.

Different from the relative simplicity of a label class condition information data structure and a comparison mode, corpus item information is sentence-level text containing semantic logic, and the matching of the corpus item information is based on a text similarity double-tower model. Therefore, in order to improve the matching accuracy and the text similarity accuracy of corpus condition items, the application can also be used as a training set for training a text similarity double-tower model by collecting matching labels in user logs. The matched label in the user log comprises a label log similar to the enterprise-side corpus in the policy-side corpus and a label log dissimilar to the enterprise-side corpus in the policy-side corpus.

For example, when the enterprise user views the matching condition of each condition item of the current policy on the policy detail page, as shown in fig. 2d, the policy detail page may have 3 condition items, and the enterprise declaration information of the user is matched with the 3 condition items of the policy file, so that the total matching degree is 100%. The policy details page is also provided with a first button and a second button, wherein the first button is a declaration button, and the second button is a modification button. When the situation that the user checks the matching condition of each condition item in the policy file on the policy detail page is detected, and the user clicks the reporting button through the policy detail page, the approval log of the enterprise user on the matching condition of each condition item under the policy is saved while the specific reporting page is skipped, so that the label log similar to the enterprise corpus at the policy end can be collected.

For the condition items which are considered to be matched and dissimilar by the user, the condition items can be changed by clicking a modification button, as shown in fig. 2e, the user modifies the declaration information of the enterprise after clicking the modification button, and then the save modification button can be clicked. As shown in fig. 2f, when the server detects that the user clicks the modification button through the policy details page, the enterprise declaration information of the user may be updated according to the modification information of the user, and then the policy details page matched with the enterprise declaration information of the user is skipped. In the embodiment of the application, after the enterprise user clicks the save and modify button, the label log which is dissimilar to the enterprise-side corpus is synchronously saved while the enterprise user information is updated.

In the embodiment of the application, after the label log data of similarity and dissimilarity of the policy corpus and the enterprise corpus are collected, the label log data can be used as a training set to train a text similarity double-tower model, so that an optimized closed loop of the whole matching process is formed.

In the embodiment of the application, the policy key information can be automatically extracted and converted into the structured data to be stored in the database. After the policy information extraction is completed, the policy structured data in the database will intelligently generate relevant questionnaires as the user browses specific policy pages. The questionnaire generating function combines the existing information of the enterprise and the condition item information of the current policy, synthesizes the weight difference between the reference label and the corpus, and generates a user questionnaire according to the information gain ratio for the user to fill in and perfect the enterprise information.

The information gain ratio determines the priority of presentation of various questions in the questionnaire generated by the enterprise. For example, the weights of the tag item and the corpus item in the matching calculation are set to be 7:3, so that when a questionnaire is generated, the problem of the tag type is preferentially displayed, and the problem of the corpus is displayed at a later position. After the user finishes or partially finishes the questionnaire, the embodiment of the application can supplement the enterprise information in the database, then uses the more perfect enterprise declaration information to carry out policy intelligent matching, and calculates the similarity between the enterprise and the policy. After the analysis of the policy file and the collection of the enterprise information through the enterprise questionnaire are completed, the embodiment of the application can calculate the similarity based on the existing information of the policy file and the enterprise information, acquire the matching degree of the policy file and the enterprise information, and recommend the matching degree to the enterprise user for display after sequencing.

After key information which can be matched with enterprises is extracted from the administrative policy files through a natural language processing model, an algorithm model is required to be built to complete calculation matching of the two information. The traditional matching method is based on the labels extracted from the policy file: and comparing and matching the value form data based on the mode of the character similarity rule. It has the following disadvantages: representing the policy file key information as labels: the value form data cannot perfectly comprise all condition information in the file, behavior description sentences existing in the policy conditions are ignored, the matching mode is also dependent on a semantic understanding model, and the matching mode of simple character similarity cannot be solved. In consideration of the requirements of matching enterprises and pushing more preferential policies, an enterprise information supplementing mechanism should be introduced in the matching process to perfect enterprise information. The traditional simple character similarity matching mode has the problems of high labor cost, low updating efficiency and the like because the rule is updated manually according to a new sample.

However, the present application implements a policy intelligent matching method, fully considering the behavior description sentences ignored by the traditional matching method, as shown in fig. 2g, the data set related to the embodiment of the present application includes policy files, policy corpus similarity text pair data sets, etc.; the policy corpus similarity text pair data set can be used for training a text similarity double-tower model; the AI model related to the embodiment of the application comprises a text similarity double-tower model, and the embodiment of the application builds the text similarity double-tower model to carry out semantic understanding matching and improve the matching degree of key information of the policy file. The policy matching link intelligently generates questionnaires to perfect enterprise information according to the policy key information and the enterprise key information, and then carries out policy intelligent matching to improve the enjoyment of enterprises on preferential policies; and then carrying out matching log collection, namely obtaining label log data of similarity and dissimilarity between the collected policy corpus and the enterprise corpus by collecting user behavior logs, and taking the label log data as a training set to optimize the iterative text similarity double-tower model, wherein the full-automatic optimization mode avoids the problems of high cost and low efficiency caused by the traditional manual optimization.

In order to better implement the above method, the embodiment of the present application further provides a policy matching device, where the policy matching device may be specifically integrated in an electronic device, and the electronic device may be a device such as a terminal, a server, or the like. The terminal can be a mobile phone, a tablet personal computer, an intelligent Bluetooth device, a notebook computer, a personal computer and other devices; the server may be a single server or a server cluster composed of a plurality of servers.

For example, in the present embodiment, a method according to an embodiment of the present application will be described in detail by taking a specific integration of a policy matching device in an electronic device as an example.

For example, as shown in fig. 3, the policy matching device may include: an information extraction module 310, an exact match module 320, a fuzzy match module 330, and a match determination module 340. The information extraction module 310 is configured to perform structured extraction on the policy file and the enterprise declaration information of the user, so as to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types; the number of different structure types includes a first structure type and a second structure type; the precise matching module 320 is configured to precisely match the policy key information of the first structure type with the enterprise key information of the first structure type, so as to obtain a precise matching degree; the fuzzy matching module 330 is configured to perform fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain a fuzzy matching degree; the matching degree determining module 340 is configured to determine the policy matching degree of the enterprise reporting information and the policy file according to the accurate matching degree and the fuzzy matching degree.

In some embodiments, the information extraction module 310 includes a policy information extraction module configured to: acquiring a policy file; performing label structured extraction on the policy file to obtain label type policy key information; and carrying out corpus structured extraction on the policy file to obtain corpus type policy key information.

In some embodiments, the information extraction module 310 includes an enterprise information extraction module configured to: acquiring enterprise declaration information of a user; carrying out label structured extraction on enterprise declaration information of a user to obtain label type enterprise key information; and carrying out corpus structured extraction on enterprise declaration information of the user to obtain corpus enterprise key information.

In some embodiments, the precision matching module 320 includes a tag class precision matching module configured to: and carrying out accurate matching on the label type policy key information and the label type enterprise key information to obtain accurate matching degree.

In some embodiments, the tag class precision matching module includes a tag class precision matching sub-module configured to: when the label type is the ith label type, acquiring a first label value corresponding to the label type enterprise key information of which the label type is the ith label type and a second label value corresponding to the label type policy key information of which the label type is the ith label type; when the ith threshold condition is met between the first label value and the second label value, the accurate matching between the label enterprise key information with the label type of the ith label type and the label policy key information with the label type of the ith label type is judged, wherein i is more than or equal to 1.

In some embodiments, the second structure type is a corpus type, and the fuzzy matching module 330 includes a fuzzy matching prediction module configured to: and carrying out fuzzy matching on the corpus policy key information and the corpus enterprise key information to obtain fuzzy matching degree.

In some embodiments, the expected class fuzzy matching module includes a fuzzy matching degree calculation module configured to: inputting corpus-type policy key information into a first BERT semantic model of a text similarity double-tower model, and outputting a policy text representation vector by the first BERT semantic model; inputting the corpus enterprise key information into a second BERT semantic model of the text similarity double-tower model, and outputting an enterprise text representation vector by the second BERT semantic model; and calculating the similarity between the policy text representation vector and the enterprise text representation vector to obtain the fuzzy matching degree.

In some embodiments, the policy matching degree includes a total matching degree between the user's business declaration information and the policy document, and a matching degree between the user's business declaration information and each condition item in the policy document; the policy matching device further comprises a sorting module and a display module, wherein the sorting module is configured in the following manner: acquiring the total matching degree between enterprise declaration information of a user and different policy files; ordering the total matching degree between the enterprise declaration information of the user and the different policy files from top to bottom to obtain a total matching degree ordering set; displaying the recommended policy to the user according to the total matching degree sequencing set; the display module is configured in: and according to the matching degree of the enterprise declaration information of the user and each condition item in the policy file, displaying the matching condition of the enterprise declaration information of the user to each condition item in the policy file on the policy detail page.

In some embodiments, the presentation module further comprises a log preservation module for: when detecting that the user checks the matching condition of each condition item in the policy file on the policy detail page, and the user clicks the first button through the policy detail page, saving an approval log; the approval log is a log approved by the user on the matching condition of each condition item of the policy file, and comprises a label log with similar policy-end corpus and enterprise-end corpus; when the fact that the user clicks the second button through the policy detail page is detected, enterprise declaration information of the user is updated, and a non-approval log is stored, wherein the non-approval log is a log which is not approved by the user for the matching condition of each condition item of the policy file, and the non-approval log comprises a label log which is dissimilar to the enterprise corpus at the policy end.

In some embodiments, the policy matching device further comprises a model training module configured to: the label logs similar to the enterprise-side corpora are used as training sets; and training the text similarity double-tower model by using the training set to obtain a trained text similarity double-tower model.

In some embodiments, the policy matching device further comprises an intelligent questionnaire generation module configured to: when the fact that the user browses the policy detail page is detected, acquiring an information gain ratio according to the policy key information and the enterprise key information; generating a user questionnaire according to an information gain ratio, wherein the information gain ratio represents the priority of each question presentation in the user questionnaire; and updating enterprise declaration information of the user through the user questionnaire.

The present invention implements a policy intelligent matching apparatus, wherein a policy tag hierarchy module (corresponding to an information extraction module 310): the category of key information in the policy file is normalized to determine the comparison mode of the result paradigm of the extracted information of the analysis link and the matching link. The intelligent questionnaire generating module: and intelligently generating a questionnaire based on the enterprise and policy information and supplementing the enterprise information. Policy intelligent matching module (corresponding to precision matching module 320 and fuzzy matching module 330): and carrying out label accurate matching and corpus fuzzy matching based on the enterprise and policy information to obtain the matching degree of the enterprise and the policy. Policy feedback optimization module (corresponding to the ranking module and the presentation module): and obtaining a condition item similarity label based on the collected user behavior log, optimizing a text similarity model, and automatically forming a closed loop.

In the implementation, each unit may be implemented as an independent entity, or may be implemented as the same entity or several entities in any combination, and the implementation of each unit may be referred to the foregoing method embodiment, which is not described herein again.

As can be seen from the above, the policy matching device of the present embodiment constructs a policy condition tag system, which includes the attribution of the types of the various condition texts in the policy file, determines the result paradigm of the policy analysis, and also determines the matching manner of the different types of conditions in the policy matching link. And then, respectively developing an accurate matching module and a fuzzy matching module according to different types of policy condition texts set in the label system design, wherein the accurate matching module and the fuzzy matching module are mainly completed in a mode of character similarity and rule comparison, and the semantic matching of the condition sentences is completed through a text similarity model. Meanwhile, the matching link comprises an intelligent questionnaire generating function, and the questionnaire is intelligently generated based on the information gain ratio by comparing information differences between enterprises and policies so as to supplement enterprise information. Finally, similar label logs and dissimilar label logs are obtained through collecting user behavior logs and used as training sets, a text similarity model is trained, and matching precision of enterprises and policies is improved.

Therefore, the embodiment of the application fully considers behavior description sentences ignored by the traditional matching mode, builds a text similarity model for semantic understanding matching, and improves the matching degree of the key information of the policy file. The matching link perfects enterprise information by intelligently generating questionnaires, and enjoyment of enterprises on preferential policies is improved. The text similarity double-tower model is optimized by collecting user behavior logs as a training set, and the problem of high cost and low efficiency caused by traditional manual optimization is avoided by a full-automatic optimization mode.

The embodiment of the application also provides electronic equipment which can be a terminal, a server and other equipment. The terminal can be a mobile phone, a tablet computer, an intelligent Bluetooth device, a notebook computer, a personal computer and the like; the server may be a single server, a server cluster composed of a plurality of servers, or the like.

In this embodiment, a detailed description will be given taking an example that the electronic device of this embodiment is a server, for example, as shown in fig. 4, which shows a schematic structural diagram of the server according to the embodiment of the present application, specifically:

The server may include one or more processors 401 of a processing core, memory 402 of one or more computer readable storage media, a power supply 403, an input module 404, and a communication module 405, among other components. Those skilled in the art will appreciate that the server architecture shown in fig. 4 is not limiting of the server and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components. Wherein:

the processor 401 is a control center of the server, connects respective portions of the entire server using various interfaces and lines, and performs various functions of the server and processes data by running or executing software programs and/or modules stored in the memory 402 and calling data stored in the memory 402, thereby performing overall monitoring of the server. In some embodiments, processor 401 may include one or more processing cores; in some embodiments, processor 401 may integrate an application processor that primarily processes operating systems, user interfaces, applications, and the like, with a modem processor that primarily processes wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 401.

The memory 402 may be used to store software programs and modules, and the processor 401 executes various functional applications and data processing by executing the software programs and modules stored in the memory 402. The memory 402 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like; the storage data area may store data created according to the use of the server, etc. In addition, memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

The server also includes a power supply 403 for powering the various components, and in some embodiments, the power supply 403 may be logically connected to the processor 401 by a power management system, such that charge, discharge, and power consumption management functions are performed by the power management system. The power supply 403 may also include one or more of any of a direct current or alternating current power supply, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.

The server may also include an input module 404, which input module 404 may be used to receive entered numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

The server may also include a communication module 405, and in some embodiments the communication module 405 may include a wireless module, through which the server may wirelessly transmit over short distances, thereby providing wireless broadband internet access to the user. For example, the communication module 405 may be used to assist a user in e-mail, browsing web pages, accessing streaming media, and so forth.

Although not shown, the server may further include a display unit or the like, which is not described herein. In this embodiment, the processor 401 in the server loads executable files corresponding to the processes of one or more application programs into the memory 402 according to the following instructions, and the processor 401 executes the application programs stored in the memory 402, so as to implement various functions in the policy matching device

In some embodiments, a computer program product is also presented, comprising a computer program or instructions which, when executed by a processor, implement the steps of any of the policy matching methods described above.

The specific implementation of each operation above may be referred to the previous embodiments, and will not be described herein.

From the above, it can be seen that the embodiment of the application can construct a text similarity model to perform semantic understanding matching, and improve the matching degree of key information of the policy file. The matching link perfects enterprise information by intelligently generating questionnaires, and enjoyment of enterprises on preferential policies is improved. The user behavior logs are collected to obtain condition item similar label logs and dissimilar label logs to be used as a training set to optimize the text similarity double-tower model, and the problem of high cost and low efficiency caused by traditional manual optimization is avoided in a full-automatic optimization mode.

Those of ordinary skill in the art will appreciate that all or a portion of the steps of the various methods of the above embodiments may be performed by instructions, or by instructions controlling associated hardware, which may be stored in a computer-readable storage medium and loaded and executed by a processor.

To this end, embodiments of the present application provide a computer readable storage medium having stored therein a plurality of instructions capable of being loaded by a processor to perform the steps of any of the policy matching methods provided by embodiments of the present application. For example, the instructions may perform the steps of: respectively carrying out structured extraction on the policy file and enterprise reporting information of the user to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types; the number of different structure types includes a first structure type and a second structure type; the policy key information of the first structure type is precisely matched with the enterprise key information of the first structure type, so that precise matching degree is obtained; fuzzy matching is carried out on the policy key information of the second structure type and the enterprise key information of the second structure type, so that fuzzy matching degree is obtained; and determining the policy matching degree of the enterprise declaration information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

Wherein the storage medium may include: read Only Memory (ROM), random access Memory (RAM, random Access Memory), magnetic or optical disk, and the like.

According to one aspect of the present application, there is provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The computer instructions are read from a computer-readable storage medium by a processor of a computer device, and executed by the processor, cause the computer device to perform the methods provided in the policy matching aspects or various alternative implementations of the policy matching aspects provided in the above-described embodiments.

The instructions stored in the storage medium may perform steps in any policy matching method provided by the embodiments of the present application, so that the beneficial effects that any policy matching method provided by the embodiments of the present application can be achieved are detailed in the previous embodiments, and are not repeated here.

The above describes in detail a policy matching method, device, server and computer readable storage medium provided by the embodiments of the present application, and specific examples are applied to illustrate the principles and embodiments of the present application, where the above description of the embodiments is only for helping to understand the method and core ideas of the present application; meanwhile, as those skilled in the art will vary in the specific embodiments and application scope according to the ideas of the present application, the present description should not be construed as limiting the present application in summary.

Claims

1. A method of policy matching, comprising:

respectively carrying out structured extraction on the policy file and enterprise reporting information of the user to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types; the number of different structure types includes a first structure type and a second structure type;

performing accurate matching on the policy key information of the first structure type and the enterprise key information of the first structure type to obtain accurate matching degree;

performing fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain fuzzy matching degree;

and determining the policy matching degree of the enterprise declaration information of the user and the policy file according to the precise matching degree and the fuzzy matching degree.

2. The policy matching method as defined in claim 1, wherein the structured extraction of the policy document and the enterprise declaration information of the user respectively obtains policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types, and the method comprises:

acquiring a policy file;

Performing label structured extraction on the policy file to obtain label type policy key information;

and carrying out corpus structured extraction on the policy file to obtain corpus type policy key information.

3. The policy matching method as defined in claim 2, wherein the structured extraction of the policy document and the enterprise declaration information of the user respectively to obtain policy key information of a plurality of different structure types and enterprise key information of a plurality of different structure types includes:

acquiring enterprise declaration information of a user;

performing label structured extraction on enterprise declaration information of the user to obtain label enterprise key information;

and carrying out corpus structured extraction on the enterprise declaration information of the user to obtain corpus enterprise key information.

4. The policy matching method according to claim 3, wherein said tag class policy key information and said tag class enterprise key information are each tagged with: the form of the label value is stored in a database; the corpus policy key information and the corpus enterprise key information are classified by the following categories: the text is stored in the database.

5. The policy matching method of claim 4, wherein said performing accurate matching on the policy key information of the first structure type and the enterprise key information of the first structure type to obtain an accurate matching degree, the first structure type being a tag type, includes:

And carrying out accurate matching on the label type policy key information and the label type enterprise key information to obtain accurate matching degree.

6. The policy matching method as defined in claim 5, wherein said performing accurate matching on said tag class policy key information and said tag class enterprise key information to obtain an accurate matching degree includes:

when the label type is the ith label type, acquiring a first label value corresponding to the label type enterprise key information of which the label type is the ith label type and a second label value corresponding to the label type policy key information of which the label type is the ith label type;

when the first label value and the second label value meet the ith threshold condition, determining that the label type is the label type i label type enterprise key information and the label type is the label type i label type policy key information are accurately matched, wherein i is greater than or equal to 1.

7. The policy matching method according to claim 4 or 6, wherein said performing fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain fuzzy matching degree, where the second structure type is a corpus type, includes:

And carrying out fuzzy matching on the corpus-class policy key information and the corpus-class enterprise key information to obtain fuzzy matching degree.

8. The policy matching method of claim 7, wherein said performing fuzzy matching between said corpus-like policy key information and said corpus-like enterprise key information to obtain fuzzy matching degree comprises:

inputting the corpus-class policy key information into a first BERT semantic model of a text similarity double-tower model, and outputting a policy text representation vector by the first BERT semantic model;

inputting the corpus enterprise key information into a second BERT semantic model of the text similarity double-tower model, and outputting an enterprise text representation vector by the second BERT semantic model;

and calculating the similarity between the policy text representation vector and the enterprise text representation vector to obtain fuzzy matching degree.

9. The policy matching method according to claim 8, wherein the policy matching degree includes a total matching degree between the user's business declaration information and the policy file, and a matching degree between the user's business declaration information and each condition item in the policy file; after determining the policy matching degree between the enterprise declaration information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree, the method further comprises the following steps:

Acquiring the total matching degree between enterprise declaration information of the user and different policy files;

sorting the total matching degree between the enterprise declaration information of the user and the different policy files according to the sequence from top to bottom to obtain a total matching degree sorting set;

displaying a recommendation policy to the user according to the total matching degree sequencing set;

and displaying the matching condition of the enterprise reporting information of the user to each condition item in the policy file on a policy detail page according to the matching degree of the enterprise reporting information of the user and each condition item in the policy file.

10. The policy matching method according to claim 9, wherein said displaying the matching condition of the user's business declaration information to each condition item in the policy file on a policy details page according to the matching degree of the user's business declaration information to each condition item in the policy file further comprises:

when detecting that the user checks the matching condition of each condition item in the policy file on the policy detail page, and the user clicks a first button through the policy detail page, saving an approval log, wherein the approval log comprises a label log with similar policy-end corpus and enterprise-end corpus;

When the user is detected to click a second button through the policy detail page, updating enterprise declaration information of the user, and storing a non-approval log, wherein the non-approval log comprises a label log with a policy end corpus dissimilar to an enterprise end corpus.

11. The policy matching method defined in claim 10, wherein said method further comprises:

the policy-end corpus and the enterprise-end corpus are similar to each other in label log, and the policy-end corpus and the enterprise-end corpus are dissimilar in label log, which is used as a training set;

and training the text similarity double-tower model by using the training set to obtain a trained text similarity double-tower model.

12. The policy matching method according to claim 1 or 11, said method further comprising:

when the user browsing the policy detail page is detected, acquiring an information gain ratio according to the policy key information and the enterprise key information;

generating a user questionnaire according to the information gain ratio, wherein the information gain ratio represents the priority of each question presentation in the user questionnaire;

and updating enterprise declaration information of the user through the user questionnaire.

13. A policy matching device, comprising:

the information extraction module is used for respectively carrying out structured extraction on the policy file and enterprise declaration information of the user to obtain a plurality of policy key information with different structure types and enterprise key information with different structure types; the number of different structure types includes a first structure type and a second structure type;

the accurate matching module is used for accurately matching the policy key information of the first structure type with the enterprise key information of the first structure type to obtain accurate matching degree;

the fuzzy matching module is used for performing fuzzy matching on the policy key information of the second structure type and the enterprise key information of the second structure type to obtain fuzzy matching degree;

and the matching degree determining module is used for determining the policy matching degree of the enterprise declaration information of the user and the policy file according to the accurate matching degree and the fuzzy matching degree.

14. An electronic device comprising a processor and a memory, the memory storing a plurality of instructions; a processor loads instructions from a memory to perform the steps in the policy matching method as claimed in any one of claims 1 to 12.

15. A computer readable storage medium having stored thereon a plurality of instructions adapted to be loaded by a processor for performing the steps of the policy matching method of any of claims 1 to 12.