File review method and device and electronic equipment

Abstract

The invention provides a file review method and device and electronic equipment, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining the feature information of a to-be-reviewed target file; generating a rule file according to a first rule based on an examination object in the feature information, wherein the rule file comprises at least one rule; and reviewing the target file by utilizing the at least one rule to obtain a reviewing result. In the embodiment of the invention, the feature information of the target file is obtained, and the rule file is generated for the review object in the feature information of the target file, so that the target file is reviewed by using the rule in the rule file, and the review efficiency of the target file can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a file review method and device and electronic equipment. BACKGROUND

[0002] In project management and business processes, file review is a key link to ensure project compliance, accuracy and quality. Currently, file review usually relies on manual operation, for example, using keyword regular matching to assist manual review. However, with the complication of business scenarios, when facing a large number of files, the efficiency of manual review is low. SUMMARY

[0003] The embodiments of the present application provide a file review method, device and electronic equipment to solve the problem of low efficiency of manual review of files.

[0004] To solve the above technical problems, the present application is implemented as follows:

[0005] In a first aspect, the embodiments of the present application provide a file review method, which comprises:

[0006] obtaining feature information of a target file to be reviewed;

[0007] generating a rule file based on a review object in the feature information according to a first rule, the rule file comprising at least one rule;

[0008] reviewing the target file using the at least one rule to obtain a review result.

[0009] In a second aspect, the embodiments of the present application provide a file review device, which comprises:

[0010] an acquisition module configured to obtain feature information of a target file to be reviewed;

[0011] a generation module configured to generate a rule file based on a review object in the feature information according to a first rule, the rule file comprising at least one rule;

[0012] a review module configured to review the target file using the at least one rule to obtain a review result.

[0013] In a third aspect, the embodiments of the present application provide an electronic equipment comprising a processor, a memory and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the file review method of the first aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the steps of the file review method in the first aspect.

[0015] In a fifth aspect, a computer program product is provided, and the computer program product includes computer instructions. The computer instructions are executed by a processor to implement the steps of the file review method in the first aspect.

[0016] In the embodiments of the present application, the feature information of the target file is acquired, the rule file is generated based on the review object in the feature information of the target file, and thus the target file is reviewed by using the rules in the rule file, so that the review efficiency of the target file can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 is one of the flowcharts of the file review method provided by the embodiments of the present application;

[0019] Figure 2 is the second flowchart of the file review method provided by the embodiments of the present application;

[0020] Figure 3 is the third flowchart of the file review method provided by the embodiments of the present application;

[0021] Figure 4 is the interface schematic diagram of the file review process provided by the embodiments of the present application;

[0022] Figure 5 is the structural schematic diagram of the file review device provided by the embodiments of the present application;

[0023] Figure 6 is the structural schematic diagram of the electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0024] The technical solutions of the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] The natural language tool can perform basic semantic analysis and keyword extraction on the text, which can be used to assist the reviewer in the review, and some predefined fixed rules can also be used to assist the reviewer in the preliminary review.

[0026] However, due to the subjectivity of the reviewer, different reviewers have different understandings and judgments on the same file, which leads to inconsistent review results, and a large number of human resources are used for review, which increases the review cost and the review period.

[0027] Embodiments of the present application provide a file review method, device and electronic equipment to solve the problem of low efficiency of manual review.

[0028] Reference Figure 1 , Figure 1 is a flowchart of a file review method provided by embodiments of the present application, as shown in Figure 1 , the method comprises the following steps:

[0029] Step 101, obtaining the feature information of the target file to be reviewed;

[0030] Step 102, generating a rule file based on the review object in the feature information according to a first rule, the rule file comprising at least one rule;

[0031] Step 103, reviewing the target file by using the at least one rule to obtain a review result.

[0032] Among them, the feature information can be the key data or attributes extracted from the target file, for example, text content, structure information, etc.

[0033] In some embodiments, the content of the target file can be parsed, the information of the target file can be extracted, and the feature information can be identified based on the extracted information by using a large model. For example, a contract file containing "R&D amount" and "project background", the extracted feature information can include "R&D amount is 5 million yuan", "project background is technology research and development".

[0034] In some embodiments, the information of the target file is extracted, and the feature information is obtained based on the information.

[0035] The review object in the feature information is the specific content that needs to be reviewed, for example, R&D amount (including specific numerical value), field, etc.

[0036] The first rule can be a user-defined rule or a system-preset rule, and can also be a rule template, or a rule generated based on user input parameters and page data, which is not limited here.

[0037] In some embodiments, the rule files can be loaded into the engine to generate a rule base. The pre-processed target file resulting business data (i.e. facts) are inserted into the engine as input for the first rule matching. For example, there is a file information extraction class to represent some elements extracted from the file to be checked. This class object can be inserted as a fact into the rule engine.

[0038] The first rules are matched according to the inserted business data (facts) to determine which rules can be triggered. When the facts of the target file are inserted, the Drools rule (i.e. rule file) engine checks which conditions of the first rules match the facts of the target file. Drools uses an algorithm (e.g. Rete algorithm) for rule matching, which can efficiently determine which rules can be triggered in a large number of rules and facts. Once the rule matching is determined, Drools will execute the matched rules to output the customized desired results. For example, pass or fail.

[0039] When loading the rule file, the “.drl” file (i.e. rule file) can be loaded into the rule engine through KieServices. Multiple first rules are generated in each rule file, but only one review result is finally obtained.

[0040] The triggered rules are executed in a certain order to process the matched facts. When the first rule includes more than two rules, the logical order of execution between these rules can be included, and the extracted data of the target file is reviewed using these rules. For example, when the priority of rule one is higher than that of rule two, rule one is used for review first, and then rule two is used for review. If the file content does not meet rule one, it means that the review fails, and it is not necessary to check whether it meets rule two, which reduces the review process and improves the review efficiency.

[0041] The above process is shown in Figure 2 Based on the target file, text recognition and extraction are performed, the recognized and extracted information is input into the AI algorithm large model, and the characteristic information of the target file is recognized by using the large model. The rule configuration platform generates a rule file based on the characteristic information, and the rules in the rule file are used to review the information of the recognized target file to obtain a review result.

[0042] Optionally, the characteristic information of the target file to be reviewed is obtained by:

[0043] The information in the target file to be reviewed is segmented to obtain document segment information;

[0044] Key information in the document segment information is recognized;

[0045] Natural language processing is performed on the key information to obtain characteristic information of the target file.

[0046] Before reviewing the target file to be reviewed, the target file is preprocessed.

[0047] For example, parse ".doc" or ".docx" files using libraries like Apache POI or docx4j, or parse PDF files using libraries like PyPDF2, PDFBox, or pdfplumber, extracting information from the file to obtain symbols such as line breaks, special characters, or formatting marks. During this process, the text is cleaned to remove unnecessary symbols and the target file's information is segmented into document fragments, such as text blocks, suitable for input into the larger model.

[0048] The document fragment information is input into the artificial intelligence (AI) big model, and the key information in the document fragment, such as the topic and summary, is read through the big model.

[0049] This process may include the following operations:

[0050] Divide document fragments into paragraphs: Understand the structure of documents, such as chapters, paragraphs, lists, tables of contents, abstracts, etc.

[0051] Named Entity Recognition (NER): Extracts important entities such as project background, R&D activity results, R&D approval amount, R&D activity scenarios, and R&D attributes.

[0052] Paragraph keyword extraction: Identify key information in the document, such as amount, percentage, amount unit (yuan, 10,000 yuan), year, project establishment, completion, design, approval, etc.

[0053] After the large model identifies the key information in the document fragment, it reorganizes the information into the required format according to needs, such as JSON, plain text, HTML, etc.

[0054] Specifically, the pre-processed text is subjected to natural language processing (NLP) and other technologies to conduct in-depth analysis of key information and paragraph information in the target document, ultimately extracting relevant information. For example, specific content such as R&D activity scenarios, R&D approval amounts, R&D investment amounts, R&D activity demonstrations, project background, R&D activity results, and design approval amounts.

[0055] Optionally, identifying key information in the document fragment information includes at least one of the following:

[0056] identify at least one key information in the topic information, entity information, abstract information, paragraph keyword, forward keyword, reverse keyword in the document segment information;

[0057] In the case of including non-text information in the document segment information, the non-text information and the position of the non-text information are identified to obtain the key information including position information;

[0058] In the case of including image information in the document segment information, the image information is converted into text information, and the text information is used to generate the key information;

[0059] In the case of including target text information in the document segment information, at least one of the following processing is performed on the target text information to obtain the key information: reorganizing a plurality of first text information in the target text information according to position sequence; filtering processing on second text information in the target text information; adding a mark to third text information in the target text information, and adding the third text information to a position in the text information indicated by the mark.

[0060] In extracting key information in the document segment information, identification of key information such as topic information, entity information, abstract information, paragraph keyword, forward keyword, reverse keyword is included.

[0061] Multi-modal information extraction: If the document contains non-text information (such as images, tables), combine the large model output (including non-text recognition results and positions) with OCR or image analysis results. The large model will additionally output document structure labels, such as recognizing and labeling non-paragraph content (headers, footers, annotations, multi-column classification), outputting structured text with position information, i.e. key information; For multi-column layout content, the large model reorders the optical character recognition (OCR) according to the reading order through coordinate information, ensuring paragraph coherence, and generating complete recognition and extraction results.

[0062] Special scenario processing - scanned PDF / OCR processing: For image information, such as PDF, use OCR to convert images to text. OCR libraries such as Tesseract will recognize each character, and then use the model to correct and correct errors to obtain the key information, to improve accuracy.

[0063] For some special scenarios, a variety of ways can be used to identify special text information (i.e. target text information). Among them, the target text information can include any one or more of the first text information, the second text information, and the third text information.

[0064] Special scene processing - complex document structure (i.e. target text information): such as multi-column layout, header and footer or annotation, etc. are processed through the processing flow of complex documents.

[0065] Among them, for the first text information, that is, the text information not arranged in the predetermined order (for example, the text information of the multi-column layout), the OCR will first output the text with text coordinates and column boundaries, and the pre-processing layer will divide the column area based on the coordinate clustering algorithm, and the large model will output the text reorganized in logical order by arranging the OCR text in physical order plus column coordinates;

[0066] For the second text information, that is, non-paragraph information, for example, header, footer and other information, the content of the suspected header will be judged and filtered by text pattern matching;

[0067] For the third text information, that is, other information not displayed in the paragraph, for example, annotations and other information, the correspondence between the annotation number (i.e. mark) and the paragraph mark (i.e. mark) will be extracted, and the annotation content will be inserted into the paragraph reference position.

[0068] In the above manner, the key information of some special text is recognized, and the recognition accuracy is improved.

[0069] Optionally, the rule file is generated according to a first rule based on the review object in the feature information, comprising:

[0070] receiving a first input of a user on the review object in the feature information;

[0071] In response to the first input, a rule file is generated using the review object according to the first rule, the rule file comprising at least one level arranged in order, each level in the at least one level comprising the review object, a comparison object, a relationship between the review object and the comparison object, and a review conclusion, the review conclusion being used to indicate the review result of the target file.

[0072] Among them, the review item configured for the review object can include the following contents:

[0073] 1) Review object: the feature information extracted from the target file using the AI large model, which can include the specific value of the review object;

[0074] 2) Relationship: logical item, >, <, =, contains, exists, consistent;

[0075] 3) Comparison object: the comparison object of the review object, which can be a system input value, a custom value, and feature information;

[0076] 4) Review conclusion and prompt information: user-defined.

[0077] Among them, the review object, relationship, and comparison object are basic information, and the information of the review object, relationship, and comparison object can be defined and entered in the platform in advance.

[0078] For example, the information of the review object is defined in the platform as "R&D amount", the relationship is "greater than", and the comparison object is "1 million yuan".

[0079] After obtaining the contract document, the "R&D Amount" feature in the contract document is extracted, which is "5 million yuan." Based on the rule (Review Object = "R&D Amount," Relationship = "Greater Than," Comparison Object = "1 million yuan"), the system determines whether "5 million yuan > 1 million yuan" is true.

[0080] The user can perform operations (ie, first input) on the visual interface to select specific feature information from the extracted feature information as the review object.

[0081] Every data and logical relationship selected when configuring the rules is ultimately assigned to the rule file through variables, allowing the user to freely define and combine multiple, multi-level review objects. When the system executes the review, it is ultimately interpreted into a standard rule file.

[0082] Before generating the rule file, you can define the first rule. The first rule includes the definition of four variables: the grammatical structure of the execution rule, the execution priority of each rule (supporting multi-level rules), the review object, the comparison object, the logical relationship, and the conclusion.

[0083] Based on the review object selected by the user, rule files are generated according to the hierarchy and priority. The user's visualized operation data is converted into the final execution file. The review object, comparison object, logical relationship, and conclusion in each hierarchy are generated into rule files, and priority calculation is performed based on the AI ​​big model to ensure the correctness of the order.

[0084] During generation, each level's review object, comparison object, logical relationship, and conclusion are converted into a rule in the rule file.

[0085] A corresponding pre-verification rule is generated for each rule. The internal logic of the verification rule is the determination of the review object, comparison object, and data type verification.

[0086] Calculate the corresponding priority of each rule when generating it to improve the accuracy of the review conclusion.

[0087] Generate a function for each logical relationship. For example, if the review object >= comparison object, you need to define a function gte to compare the two parameters and determine the review result.

[0088] When there are multiple rules, a hierarchy can be set according to the order of review, for example, rule one is the first level, rule two is the second level, and rule three is the third level, each level includes information such as review object, comparison object, relationship, review conclusion, etc. When reviewing these rules, the characteristic information of the target file is reviewed according to the order of the corresponding rules of the hierarchy.

[0089] The user can configure through the rule configuration platform (as shown in Figure 3 The rule supports the following ways:

[0090] 1. Basic verification rule, check if the data conforms to the predetermined format or logic;

[0091] 2. Data extraction rule, extract specific entities such as amount and date from unstructured text;

[0092] 3. Conditional branching rule, perform different operations according to the results of the pre-check rule;

[0093] 4. Custom prompt, define clear prompts for success and failure scenarios. Support multi-level review rule configuration, implicit review rule configuration (pre-check rule is an implicit rule transparent to the user).

[0094] As shown in Figure 3 , set rule one and rule two based on review object one (for any review object), and set pre-check rule. For each rule, if the pre-check rule fails, output the fail prompt; if the pre-check rule passes, further review whether it meets rule one (or rule two), if it does not meet rule one (or rule two), output the fail prompt. If the pre-check rule, rule one and rule two all pass, output the pass prompt.

[0095] In this way, the user can customize the review object, comparison object and logical relationship according to the specific business requirements to adapt to the audit requirements of different scenarios. And through the hierarchical structure design, it supports multi-level condition combination (such as "first check format, then check data"), which meets the complex business logic.

[0096] The visualization part of the platform can include the configuration of basic information. The review conclusion and prompt information can be customized by the user when configuring the review object. The user can configure through the rule configuration platform, and get the review result through the visualization platform, the interface can display the content as shown in Figure 4 .

[0097] Optionally, the rule file is generated according to the first rule using the review object, including:

[0098] receiving a first parameter input by a user;

[0099] generating the first rule based on the review object and the first parameter, the first rule including a placeholder of an editable parameter;

[0100] receiving a second parameter input by the user based on the placeholder;

[0101] generating the rule file based on the second parameter and the first rule.

[0102] For rules with similar logic but different input data, a user can define a first rule by inputting a first parameter in a visual interface. The generated first rule includes a placeholder of an editable parameter, and the user inputs a second parameter in the placeholder, and generates a rule file based on the second parameter and the first rule.

[0103] That is, the first rule is used to define the skeleton of the rule, and includes a placeholder of a variable parameter. Based on the first rule, the rule file is generated in combination with the second parameter input by the user.

[0104] The first rule can use FreeMarker syntax to represent the parameter.

[0105] An example file template.drt is as follows:

[0106]

[0107] When the input data changes, only the first rule or dynamic data needs to be updated to quickly generate a new rule file without modifying the existing rule code. This can avoid repeatedly writing rules with the same logic, reduce repetitive code, and enhance flexibility. The above-described method achieves flexibility in rule configuration, and improves review efficiency and accuracy.

[0108] Optionally, the generating the rule file based on the first rule and the review object comprises at least one of the following:

[0109] generating the rule file based on a verification rule configured by input of the user and the review object;

[0110] generating the rule file based on a business rule generated based on a third parameter input by the user and business logic and the review object;

[0111] generating a verification rule for an unexplicitly configured parameter, and generating the rule file based on the verification rule and the review object.

[0112] In some embodiments, the user inputs according to the business scenario to configure the built-in check rule, and generates a Drools Rule Language (DRL) file based on the built-in check rule and the review object;

[0113] In some embodiments, a custom rule is generated based on the third parameter (including review parameters, etc.) input by the user on the page and the business logic (i.e., business information), and a rule file is generated based on the custom rule and the review object. The review parameters can include review objects, comparison objects, and other information.

[0114] In this way, the rule file can be dynamically generated based on the parameters input by the user and the business information.

[0115] For parameters that are not displayed (not explicitly configured), such as null value checking, data type checking (such as numerical value, date, etc.), built-in check rules are added. The rule file is generated using the check rule and the review object.

[0116] Through dynamic generation of rule files, user-defined check rules, automatic supplement of check logic, etc., flexibility, automation, and efficiency of rule configuration are achieved, manual costs are reduced, and audit efficiency and data quality are improved. The completeness and accuracy of the rules are improved, so that the user only needs to focus on the business process and results when configuring the rules on the page, without considering the check problem.

[0117] Optionally, the rule file includes at least two rules, and the review of the target file using the at least one rule to obtain a review result includes:

[0118] Obtaining the execution priority of the at least two rules;

[0119] According to the execution priority, the at least two rules are used in turn to review the target file to obtain a review result, wherein the execution priority includes any of the following:

[0120] A first priority set in advance;

[0121] A second priority set based on a large model;

[0122] A third priority obtained by the user operating and setting the first priority or the second priority;

[0123] A fourth priority determined according to the level of the rule set by the user;

[0124] A fifth priority set according to business data.

[0125] The execution priority of the at least two rules can represent the order of rule review.

[0126] The priority can be set in any of the above manners, and the priority can also be dynamically adjusted according to the priority setting manner.

[0127] In some embodiments, the first priority can be pre-set, for example, the first priority pre-set according to a rule type system, or the first priority set by user operation.

[0128] In some embodiments, in the rule engine, the execution order of the rules (i.e., the execution priority) can be set as default to be uncertain, the execution order of the rules is optimized based on an AI large model to obtain a second priority; the second priority can also be dynamically adjusted on the basis of the pre-set first priority; in addition, the second priority can also be automatically adjusted according to user's demand and business data and the like information.

[0129] In some embodiments, the third priority can be dynamically adjusted on the basis of the first priority or the second priority according to user's demand, and the execution order can be intelligently recommended. In some embodiments, on the basis of the second priority optimized and adjusted by the large model, the user can define the execution order of the rules explicitly by adjusting the salience attribute.

[0130] When there are multiple rules, the execution order of the rules can be controlled by setting the salience attribute of each rule. Among them, the rule with a higher salience value will be executed before the rule with a lower salience value; if the salience values of two rules are the same, the execution order of the two rules will depend on the order defined in the Drools file.

[0131] The following is a specific example of adjusting the execution order by setting the salience attribute of the rule.

[0132] rule "valid rule" / / rule for checking passed

[0133] salience 10

[0134] end

[0135] rule "invalid rule" / / rule for checking failed

[0136] salience 10

[0137] End

[0138] rule "rule 1 rule" / / rule for checking passed

[0139] salience 9

[0140] End

[0141] rule"!Rule 1rule" / / Rule 1 failed

[0142] salience 9

[0143] End

[0144] rule"valid rule2" / / rules that passed the second-level rule verification

[0145] salience 8

[0146] end

[0147] rule"!valid rule2" / / Rule that failed the second-level validation

[0148] salience 8

[0149] End

[0150] In the above example, "valid rule" and "!valid rule" will be executed first because their salience value is 10, the highest. Then "rule1rule" and "!rule1rule" will be executed with a salience of 9. Finally, "validrule2" and "!valid rule2" will be executed with a salience of 8.

[0151] For the above scenario, when dynamically generating rules in DRL, salience can be dynamically calculated based on the user-configured rule hierarchy, and the fourth priority can be determined based on the salience value. In addition, the priority of the built-in validation rules can be set higher.

[0152] During actual execution, the execution priority is set in static order and dynamic order.

[0153] Static order: The first priority is predefined by salience or physical order in the rule file.

[0154] Dynamic order: In real-world business scenarios, the importance of rules may vary depending on context (such as user needs and data characteristics), necessitating dynamic order optimization. Based on the statically defined first-priority order, the first priority is adjusted based on business data characteristics or user needs to arrive at a fifth priority. While the static order serves as a baseline, dynamic adjustments can supplement the need to address complexity and real-time requirements.

[0155] The user can explicitly define the execution order of the rules by adjusting the salience attribute according to the recommended execution order. The rule execution level optimization is implemented, and the execution efficiency of the overall rules is improved.

[0156] By the above priority setting mode, the AI large model is used to pre-train the rule execution order, the rule execution level optimization is implemented, and the execution efficiency of the overall rules is improved. Moreover, the execution priority output by the pre-trained AI model is more in line with the review rules customized by the user and the expected review result.

[0157] By defining the rule engine template (DRL), dynamically generating the rule file (.drl), and dynamically judging the execution priority, the accuracy of the result of the multi-level rules can be improved. According to the page configuration, the condition left side (LHS) of the triggered rule is dynamically generated, and according to the expected review result of the user, the operation right side (RHS) executed when the condition is met is dynamically configured, so that the individual needs of the user can be met.

[0158] The present application can configure the individual review object and the multi-level review rule through the visual operation mode of the rule configuration platform, so that the code development of the user is reduced. For a new review object, the dynamic adjustment can be performed on the basis of the defined review rule, the program development is reduced, the flexibility of the system adjustment is improved, and the human cost is saved.

[0159] Referring to Figure 5 , Figure 5 is a structural schematic diagram of a file review device provided by an embodiment of the present application, as Figure 5 indicated, the file review device 500 includes:

[0160] The acquisition module 501 is configured to acquire feature information of a target file to be reviewed.

[0161] The generation module 502 is configured to generate a rule file according to a first rule based on a review object in the feature information, and the rule file includes at least one rule.

[0162] The review module 503 is configured to review the target file by using the at least one rule, and obtain a review result.

[0163] Optionally, the generation module 502 includes:

[0164] The first receiving sub-module is configured to receive a first input of a user on the review object in the feature information.

[0165] The first generating sub-module is configured to generate a rule file according to the first rule and based on the review object in response to the first input, the rule file including at least one level arranged in sequence, each level of the at least one level including the review object, a comparison object, a relationship between the review object and the comparison object, and a review conclusion, the review conclusion being configured to indicate a review result of the target file.

[0166] Optionally, the generating module 502 includes:

[0167] The second receiving sub-module is configured to receive a first parameter input by a user.

[0168] The second generating sub-module is configured to generate the first rule based on the review object and the first parameter, the first rule including a placeholder of an editable parameter.

[0169] The third receiving sub-module is configured to receive a second parameter input by the user based on the placeholder.

[0170] The fourth generating sub-module is configured to generate the rule file based on the second parameter and the first rule.

[0171] Optionally, the first generating sub-module is configured to perform at least one of the following:

[0172] Generate the rule file based on a verification rule configured based on an input of the user and the review object.

[0173] Generate the rule file based on a business rule generated based on a third parameter input by the user and a business logic and the review object.

[0174] Generate a verification rule for an unexplicitly configured parameter, and generate the rule file based on the verification rule and the review object.

[0175] Optionally, the rule file includes at least two rules, and the review module 503 includes:

[0176] The obtaining sub-module is configured to obtain an execution priority of the at least two rules.

[0177] The review sub-module is configured to sequentially review the target file based on the execution priority and using the at least two rules to obtain a review result, wherein the execution priority includes any of the following:

[0178] A first priority set in advance.

[0179] A second priority set based on a large model.

[0180] A third priority obtained by the user operating and setting the first priority or the second priority.

[0181] a fourth priority determined according to a level in which the user setting rule is located;

[0182] a fifth priority set according to business data.

[0183] Optionally, the obtaining module comprises:

[0184] a splitting sub-module, configured to split information in the target file to be reviewed to obtain document segment information;

[0185] a recognizing sub-module, configured to recognize key information in the document segment information;

[0186] a processing sub-module, configured to perform natural language processing on the key information to obtain feature information of the target file.

[0187] Optionally, the recognizing sub-module is specifically configured to perform at least one of the following:

[0188] recognize at least one of theme information, entity information, abstract information, paragraph keywords, positive keywords and reverse keywords in the document segment information;

[0189] in a case where the document segment information comprises non-text information, recognize the non-text information and a position of the non-text information to obtain the key information comprising position information;

[0190] in a case where the document segment information comprises image information, convert the image information into text information, and generate the key information by using the text information;

[0191] in a case where the document segment information comprises target text information, perform at least one of the following on the target text information to obtain the key information: reorganize a plurality of first text information in the target text information according to a position sequence; filter a second text information in the target text information; add a mark to a third text information in the target text information, and add the third text information to a position of text information indicated by the mark.

[0192] The file review device can implement each process implemented in the method embodiment and achieve the same technical effects. To avoid repetition, details are not described herein. Figure 1

[0193] It should be noted that the electronic device provided in the embodiments of the present application can execute the file review device described above, and all implementation manners in the file review method embodiments are applicable to the electronic device, and can achieve the same or similar beneficial effects. To avoid repeated description, details are not described herein.​

[0194] The embodiment of the present application also provides a terminal. Since the principle of the terminal to solve the problem is similar to the file review method in the embodiment of the present application, the implementation of the terminal can refer to the implementation of the method, and the repeated parts will not be repeated. Figure 6 As shown, the terminal of the embodiment of the present application includes: a processor 600, which is used to read the program in the memory 620 and perform the following processes:

[0195] Obtain characteristic information of the target file to be reviewed;

[0196] generating a rule file according to a first rule based on the review object in the feature information, wherein the rule file includes at least one rule;

[0197] The target file is reviewed using the at least one rule to obtain a review result.

[0198] The transceiver 610 is configured to receive and send data under the control of the processor 600 .

[0199] Among them, Figure 6 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 600 and memory represented by memory 620. The bus architecture may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 610 may be a plurality of components, i.e., a transmitter and a transceiver, providing a unit for communicating with various other devices on a transmission medium. The processor 600 is responsible for managing the bus architecture and general processing, and the memory 620 may store data used by the processor 600 when performing operations.

[0200] Optionally, the processor 600 is further configured to read a program in the memory 620 and execute the steps of obtaining characteristic information of the target file to be reviewed, including:

[0201] Segment the information in the target document to be reviewed to obtain document fragment information;

[0202] Identifying key information in the document fragment information;

[0203] Natural language processing is performed on the key information to obtain characteristic information of the target file.

[0204] Optionally, the processor 600 is further configured to read a program in the memory 620 to execute the identifying of key information in the document fragment information, including at least one of the following:

[0205] identify at least one key information in the document segment information, wherein the at least one key information comprises at least one of topic information, entity information, abstract information, paragraph keyword, forward keyword, and reverse keyword;

[0206] In a case where the document segment information comprises non-text information, identify the non-text information and a position of the non-text information, and obtain the key information comprising position information;

[0207] In a case where the document segment information comprises image information, convert the image information into text information, and generate the key information by using the text information;

[0208] In a case where the document segment information comprises target text information, perform at least one of the following processing on the target text information to obtain the key information: reorganize a plurality of first text information in the target text information according to a position sequence; filter a second text information in the target text information; add a third text information in the target text information to a position of text information indicated by a mark.

[0209] Optionally, the processor 600 is further configured to read a program in the memory 620, and perform the generating, according to the first rule, of the rule file by using the review object, comprising:

[0210] receiving a first input of a user on the review object in the feature information;

[0211] in response to the first input, generating, according to the first rule, the rule file by using the review object, the rule file comprising at least one level arranged in sequence, each level in the at least one level comprising the review object, a comparison object, a relationship between the review object and the comparison object, and a review conclusion, the review conclusion being used to indicate a review result of the target file.

[0212] Optionally, the processor 600 is further configured to read a program in the memory 620, and perform the generating, according to the first rule, of the rule file by using the review object, comprising:

[0213] receiving a first parameter input by a user;

[0214] generating the first rule based on the review object and the first parameter, the first rule comprising a placeholder of an editable parameter;

[0215] receiving a second parameter input by a user based on the placeholder;

[0216] generating the rule file based on the second parameter and the first rule.

[0217] Optionally, the processor 600 is further configured to read a program in the memory 620 and execute the process of generating a rule file according to a first rule based on the review object in the feature information, including at least one of the following:

[0218] Based on the verification rules configured by the user's input, the rule file is generated using the review object;

[0219] Generate the rule file using the review object based on the third parameter input by the user and the business rules generated by the business logic;

[0220] A verification rule for the non-explicitly configured parameter is generated, and the rule file is generated using the review object based on the verification rule.

[0221] Optionally, the rule file includes at least two rules, and the processor 600 is further configured to read a program in the memory 620 and execute the program to review the target file using the at least one rule to obtain a review result, including:

[0222] Obtaining execution priorities of the at least two rules;

[0223] According to the execution priority, the target file is reviewed in sequence using the at least two rules to obtain a review result, wherein the execution priority includes any one of the following:

[0224] Pre-set first priority;

[0225] Second priority based on large model settings;

[0226] a third priority level obtained by a user operating and setting the first priority level or the second priority level;

[0227] The fourth priority is determined by the level at which the user sets the rule;

[0228] The fifth priority is set based on business data.

[0229] The terminal provided in the embodiment of the present application can execute the above method embodiment, and its implementation principle and technical effects are similar, which will not be repeated in this embodiment.

[0230] The present application also provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program implements the various processes of the document review method embodiment described above and achieves the same technical effects. To avoid repetition, the details are not described here. The computer-readable storage medium may be, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0231] The embodiment of the present application also provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the above-mentioned Figure 1 The various processes of the method embodiment shown in the above are implemented by the computer program product, and the same technical effects can be achieved, so no further description is given here.

[0232] It should be noted that in this document, the term "comprising" or "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0233] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present application.

[0234] The embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the above-mentioned specific embodiments, the above-mentioned specific embodiments are only illustrative, not restrictive, and those skilled in the art can make many forms under the inspiration of the present application without departing from the scope of the present application and the protection scope of the claims.

Claims

1. A document review method, characterized in that: include: Obtain characteristic information of the target file to be reviewed; generating a rule file according to a first rule based on the review object in the feature information, wherein the rule file includes at least one rule; The target file is reviewed using the at least one rule to obtain a review result.

2. The method according to claim 1, characterized in that The generating of a rule file according to a first rule based on the review object in the feature information includes: receiving a first input from a user regarding a review object in the feature information; In response to the first input, a rule file is generated using the review object according to the first rule, and the rule file includes at least one level arranged in sequence, each level in the at least one level includes the review object, the comparison object, the relationship between the review object and the comparison object, and the review conclusion, and the review conclusion is used to indicate the review result of the target file.

3. The method according to claim 2, characterized in that Generating a rule file using the review object according to the first rule includes: Receive a first parameter input by a user; generating the first rule based on the review object and the first parameter, wherein the first rule includes a placeholder for an editable parameter; receiving a second parameter input by a user based on the placeholder; The rule file is generated based on the second parameter and the first rule.

4. The method according to claim 2, characterized in that The step of generating a rule file using the review object according to the first rule includes at least one of the following: Based on the verification rules configured by the user's input, the rule file is generated using the review object; Generate the rule file using the review object based on the third parameter input by the user and the business rules generated by the business logic; A verification rule for the non-explicitly configured parameter is generated, and the rule file is generated using the review object based on the verification rule.

5. The method according to any one of claims 1 to 4, characterized in that The rule file includes at least two rules, and the target file is reviewed using the at least one rule to obtain a review result, including: Obtaining execution priorities of the at least two rules; According to the execution priority, the target file is reviewed in sequence using the at least two rules to obtain a review result, wherein the execution priority includes any one of the following: Pre-set first priority; Second priority based on large model settings; a third priority level obtained by a user operating and setting the first priority level or the second priority level; The fourth priority is determined by the level at which the user sets the rule; The fifth priority is set based on business data.

6. The method according to any one of claims 1 to 4, characterized in that The step of obtaining characteristic information of the target file to be reviewed includes: Segment the information in the target document to be reviewed to obtain document fragment information; Identifying key information in the document fragment information; Natural language processing is performed on the key information to obtain characteristic information of the target file.

7. The method according to claim 6, characterized in that The identifying key information in the document fragment information includes at least one of the following: Identifying at least one key information among subject information, entity information, summary information, paragraph keywords, forward keywords, and reverse keywords in the document fragment information; In the case where the document fragment information includes non-text information, identifying the non-text information and the position of the non-text information to obtain the key information including the position information; In the case where the document fragment information includes image information, converting the image information into text information, and generating the key information using the text information; In a case where the document fragment information includes target text information, performing at least one of the following processing on the target text information to obtain the key information: reorganizing a plurality of first text information in the target text information according to a position order; filtering a second text information in the target text information; A mark is added to the third text information in the target text information, and the third text information is added to the position in the text information indicated by the mark.

8. A document review device, characterized in that: include: An acquisition module is used to obtain characteristic information of the target file to be reviewed; a generating module, configured to generate a rule file according to a first rule based on the review object in the feature information, wherein the rule file includes at least one rule; The review module is used to review the target file using the at least one rule to obtain a review result.

9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the document review method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the document review method according to any one of claims 1 to 7.

11. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the document review method according to any one of claims 1 to 7.

Images