Intelligent fire file examination method based on semantic comprehension and rule engine

Abstract

The invention provides an intelligent fire file review method based on semantic comprehension and a rule engine, and relates to the technical field of administrative data processing. The system is specially used for solving the problem of intelligent review of fire accident investigation files, fills the technical blank field of intelligent review of fire accident investigation files, constructs a file review standard knowledge graph, designs an intelligent review engine, and realizes full-process automation from original files to intelligent review through the intelligent file review system. The paper marking process can be monitored in real time, and cooperation of evaluation results is facilitated; the manual evaluation time cost is reduced, and the fire investigation file quality supervision and evaluation efficiency and frequency of a superior fire rescue institution for a fire investigation department are improved; primary-level personnel can find the problems of the file in time before the fire accident investigation file is rolled, then the file content is supplemented and perfected, the legal consciousness and skill level of the primary-level fire investigation personnel are improved, and the primary-level fire investigation level is promoted to be developed in the evidence, programming and technical direction.

Description

Technical Field

[0001] This invention relates to the field of administrative data processing technology, and in particular to an intelligent review method for fire case files based on semantic understanding and rule engine. Background Technology

[0002] A typical fire accident investigation file generally includes: a fire accident determination report (a formal document outlining the basic facts of the case, the determined time of ignition, location of the fire (ignition point), cause of the fire, and related evidence); a fire scene investigation report (a written record of the on-site investigation); interrogation records (records of interviews with relevant personnel); on-site photos / maps; fire loss statistics; and inspection and appraisal opinions, etc. The fire accident determination report issued by the fire and rescue agency serves as crucial evidence and is subject to cross-examination by both parties in civil, criminal, and administrative litigation proceedings. Cross-examination will focus on the legality, authenticity, and relevance of the fire accident investigation conclusions recorded in the fire accident determination report, specifically including whether the fire facts are clear, whether the evidence is sufficient and conclusive, and whether it complies with legal procedures. Its primary form is the fire accident investigation file. Therefore, the fire accident investigation file is particularly important, serving as legal proof of the fire facts; its completeness and authority are the foundation of the fire accident investigation work and an important basis for subsequent civil, criminal, and administrative litigation; its recorded content must withstand the test of history and law.

[0003] Fire accident investigation is a legally mandated administrative matter, and case file review involves a comprehensive examination of the aforementioned fire accident investigation files. Currently, to promote the standardization and scientific development of fire accident investigation and handling, and to comprehensively evaluate the effectiveness of the professional talent pool for fire accident investigation, each fire and rescue brigade and detachment regularly conducts fire accident investigation case file reviews. These reviews adhere to the "Rules for Reviewing Fire Accident Investigation Files" and are conducted in accordance with the filing requirements of each province, unifying the review standards. The review content is divided into basic elements and general elements. Basic elements focus on reviewing whether there are serious enforcement problems in terms of the subject, factual evidence, application of laws and regulations, and enforcement procedures. General elements focus on reviewing whether there are general problems in terms of document preparation, enforcement procedures, information technology application, and archiving. The quality of the fire accident investigation is comprehensively judged in conjunction with the substantive findings. The review implements a verification system, with at least one person conducting the evaluation and one person conducting the verification. For each reviewed case file, the reviewers must complete an evaluation record form, listing the details of the evaluation issues, deductions, and the final score. For example, the fire and rescue brigade and its branches regularly conduct centralized review and evaluation of case files, inviting senior professionals in the field to explain in detail the precautions in case file preparation, assessing the quality of case files and the credibility of investigation conclusions based on review rules and relevant requirements, scoring case files and selecting outstanding case files; a fire and rescue brigade held a centralized review meeting for fire accident investigation case files throughout the city, where participants discussed and rectified outstanding issues in the case files through centralized review and mutual evaluation.

[0004] Traditional fire accident investigation case file reviews mainly have the following problems: The review of case files is inefficient: Traditional fire accident investigation case files rely on manual review by reviewers. Due to the lengthy and relatively inefficient manual review process, a single fire accident investigation case file containing dozens or even hundreds of pages of evidence often takes 2 to 5 hours to complete the review.

[0005] Due to the subjective influence of the evaluators, the evaluation standards are not entirely uniform: fire accident investigation files are highly specialized and lengthy (usually hundreds of pages), making them heavily reliant on the expertise of the evaluators. Currently, although each province has unified evaluation rules, different evaluators do not have completely consistent focus points and judgment criteria, leading to omissions or discrepancies in judgments and a lack of objectivity and consistency during the evaluation process, thus affecting the evaluation results and scores. Furthermore, it is difficult to ensure strict consistency in the evaluation results for the same evaluator each time.

[0006] Limited real-time monitoring and collaborative review results, coupled with a lack of intelligent support tools: Reviewers are required to complete review record forms after reviewing case files, listing details of review issues, deductions, and final scores. This forces them to record issues and their corresponding page numbers, as well as deductions and final scores, all while lacking intelligent tools for issue organization and score calculation. Furthermore, for more complex cases, discussions are typically conducted offline or online, but these discussions lack intelligent tools for collaboratively presenting case file content and review results.

[0007] High labor costs and difficulty in increasing the frequency of reviews: Reviewing fire accident investigation files is a crucial operational activity for promptly identifying and addressing deficiencies, thereby improving file quality and investigation standards. Manual review requires a significant investment of manpower, necessitating dedicated online (or offline) meetings for focused evaluation. This often consumes substantial time from multiple departments and participants, increasing labor costs and hindering the frequency of reviews, thus preventing effective evaluation of the development of a professional fire accident investigation team.

[0008] Currently, intelligent review of fire accident investigation files is still a technological gap, and there are no specific technologies or methods in China for intelligent review of fire accident investigation files. Existing intelligent review technologies are mainly applied in the field of legal contracts or general contract management. For example, the invention patent with patent application number CN202210917535.X, "An Intelligent Contract Review Method Based on Natural Language Understanding," uses a fixed and unified standard to automatically review various contract risks according to legal requirements and the needs of enterprise operations. Alibaba Cloud's Tongyi FaRui Big Model, deeply rooted in the legal field, integrates advanced technologies such as natural language processing, machine learning, and big data analysis. It helps legal professionals efficiently manage legal documents, analyze case materials, provide intelligent legal consultation services, generate legal documents, and perform case retrieval and analysis, thereby improving the efficiency and accuracy of legal case handling. The invention patent with application number CN202211115790.9, "An Intelligent Contract Review Method and Device," is used for contract management in power companies. Based on a contract element library and a contract review rule library, it constructs a contract review model, which automatically reviews target contracts, enabling rapid contract review and improving efficiency. However, these intelligent review technologies are not specifically designed for fire accident investigation files and cannot address the need for reviewing fire accident investigation files. Summary of the Invention

[0009] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing an intelligent review method for fire case files based on semantic understanding and rule engine, constructing a standard knowledge graph for case file review, designing an intelligent review engine, and achieving full-process automation from original case files to intelligent review through intelligent case file review.

[0010] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: On the one hand, this invention provides a method for intelligent review of fire case files based on semantic understanding and rule engine, including the following steps: Step 1: During the initial review of fire accident investigation files, the fire and rescue brigade or battalion shall formulate file review standards in accordance with the "Rules for Reviewing Fire Accident Investigation Files" and the file compilation requirements of each province, and form a knowledge graph of file review standards for intelligent review; if it is not the first review, skip directly to Step 2. Step 2: Subsequent review uses an intelligent review engine to conduct intelligent review of the case files; the intelligent review engine automatically performs different review operations based on the different review points involved in each review checklist; based on the review results, a preliminary review report is generated, including a structured review report, review score, and suggestions for supplementing evidence and making revisions; Step 3: Manual intervention; the marking experts review the generated review report, confirm the automatic review results, and verify the identified problems and passed items; they manually check and fine-tune the elements marked as "uncertain to extract"; for problems that cannot be handled and require subjective judgment, they supplement the content of the review report. Step 4: After manual verification and confirmation that there are no errors, generate a case file review report.

[0011] Furthermore, in step 1, the specific method for forming the knowledge graph of case review standards for intelligent review is as follows: Step 1.1: Manually sort out the content and requirements of case file review, transform the case file review rules stipulated in the provisions into review points, assign them corresponding scores, and form a list of case file review rules for intelligent review; Step 1.2: Deconstruct and break down the case file review points, extract and create a structured case file review element library, and establish a mapping relationship between case file review rules, case file review dictionary, case file review elements and judgment logic; Step 1.3: Construct a knowledge graph of case review standards for intelligent review. The knowledge graph includes all case review rules and their corresponding scores, review dictionary, review elements and judgment logic; Step 1.4: Synchronize the case file review standard knowledge graph of intelligent review with the algorithm engineer. Based on the constructed case file review standard knowledge graph, the algorithm engineer uses graph traversal and entity mapping algorithms to convert the structured review elements, review dictionary constraints, and judgment logic into dynamic scripting language code logic that the rule engine can execute. Relying on large model technology, the engineer writes review rules to cover the intelligent review rule system of all review points. At the same time, based on the actual effect of intelligent review feedback from the training set case files, the engineer dynamically adjusts the knowledge graph information corresponding to the review rules, continuously fine-tunes and updates the case file review standard knowledge graph, and finally forms an iterative intelligent review standard knowledge graph and an intelligent review engine that can be directly called.

[0012] Furthermore, the specific method of step 2 is as follows: Step 2.1: The fire and rescue brigade or battalion uploads the fire accident investigation file from the client and uses the default review list to conduct intelligent review of the uploaded file; however, experts from the brigade or battalion are allowed to select the required review points from the default review list and customize a unified list of file review rules. Step 2.2: Utilize the intelligent review engine to conduct an intelligent review of the case file; specifically including: Step 2.2.1: Document format and structure analysis; The uploaded case files are formatted and converted into readable text; natural language processing technology is used to identify document types and areas such as titles, paragraphs, tables, seals, and signatures to understand the logical structure of the documents. Step 2.2.2: Structured information extraction and feature matching; Based on the case file review element library defined in the selected case file review rule list, structured information is automatically extracted from the parsed case file text to match predefined review elements and dictionaries; Based on rule-based automatic review, each review rule is logically compared with the extracted review elements; if the logical comparison of the review rule passes, full marks are awarded; if it fails, all marks for the corresponding review rule are deducted. Step 2.2.3: Summarize the review results of all rules, give the review conclusion of each review rule based on the review results, and calculate the score result of this review in combination with the weight; Step 2.2.4: Automatically generate case file summary; Understand the content of the documents, summarize the main contents of the case file, and generate a case file summary; Step 2.2.5: Automatically identify logical contradictions in the statements within the case file; Understand the statements in the interrogation record and, by combining them with the factual basis in the fire accident investigation report, identify any contradictions that contradict the established facts; Step 2.2.6: Based on the output of Steps 2.2.3 to 2.2.5, automatically generate a preliminary review report. This report includes a summary of the case file content, logical contradictions, review conclusions of the case file review rules, suggestions for supplementing evidence (revision opinions), and scoring results.

[0013] Furthermore, the identification process in step 2.2.1 includes optical character recognition, signature recognition, and seal recognition; Optical character recognition (OCR) is used to perform full-text recognition of case files based on PaddlePaddle optical character recognition technology, obtaining the text content and corresponding confidence level of each document in the case file. Signature recognition utilizes OpenCV image processing algorithms to extract handwritten signature regions, identifies the presence of a signature through feature matching, and assesses the quality and consistency of the signature. Seal recognition uses OpenCV image processing algorithms for seal detection and comparison; it automatically locates the red circular seal on the document and performs feature matching with a standard seal template to verify the seal type and authenticity.

[0014] Furthermore, step 2.2.2 specifically includes a document structuring process and an element extraction process; During the document structuring process, the results of optical character recognition are analyzed. Based on predefined document title keywords and layout rules, the case files are divided according to document type to form independent document units. The content of each document is then structured, analyzed, and annotated. During the element extraction process, based on natural language processing and regular rules, text elements and image elements are extracted and reviewed from the structured document content. Text elements include time, place, people, reasons, and behaviors; image elements include handwritten signatures, seal images, location indicator images, and on-site photos.

[0015] On the other hand, this application proposes an electronic device, including: one or more processors, and a memory for storing instructions, which, when executed by the one or more processors, cause the one or more processors to execute the intelligent fire case file review method based on semantic understanding and rule engine.

[0016] Fourthly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the aforementioned intelligent fire case file review method based on semantic understanding and a rule engine.

[0017] Fifthly, this application proposes a computer program product, including a computer program or instructions that, when executed by a processor, implement the aforementioned intelligent fire case file review method based on semantic understanding and rule engine.

[0018] The beneficial effects of adopting the above technical solution are as follows: The intelligent review method for fire case files based on semantic understanding and rule engine provided by this invention is specifically designed to solve the intelligent review of fire accident investigation files, filling the technical gap in the intelligent review of fire accident investigation files. It can unify the standards for reviewing fire accident investigation files, ensure the objectivity and consistency of review results, and reduce the impact of omissions or differences in judgment on review results and scores. Using the intelligent auxiliary tools of this invention, the review process can be monitored in real time, facilitating collaborative review results; it can reduce the time cost of manual review, improve the efficiency and frequency of supervision and evaluation of the quality of fire investigation files by higher-level fire and rescue agencies; it enables grassroots personnel to promptly identify problems in fire accident investigation files before they are compiled, thereby supplementing and improving the content of the files, enhancing the legal awareness and skill level of grassroots fire investigators, and promoting the development of grassroots fire investigation towards evidence-based, procedural, and technical approaches. Attached Figure Description

[0019] Figure 1 A flowchart of the intelligent fire case file review method provided in the first embodiment of the present invention; Figure 2 The flowchart for constructing a knowledge graph of case file review standards provided in the first embodiment of the present invention; Figure 3 A flowchart of the knowledge graph for fine-tuning and updating case file review standards provided in the first embodiment of the present invention; Figure 4The flowchart for constructing an intelligent review engine is provided for the first embodiment of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0021] Example 1

[0022] A method for intelligent review of fire case files based on semantic understanding and rule engine, such as Figure 1 As shown, the method of this embodiment is described below.

[0023] Step 1: During the initial review of fire accident investigation files, the fire brigade or battalion, in accordance with the "Rules for Reviewing Fire Accident Investigation Files" and the filing requirements of each province, formulates file review standards and forms a knowledge graph of file review standards for intelligent review. If this is not the initial review, proceed directly to Step 2. Figure 2 As shown, the initial review process for forming a knowledge graph of case review standards for intelligent review specifically includes the following steps: Step 1.1: Manually sort out the content and requirements of case file review, transform the case file review rules stipulated in the provisions into review points, assign them corresponding scores, and form a list of case file review rules for intelligent review.

[0024] For example, in the review rules for "enforcement procedures", the review point includes "no explanation before the fire accident determination is implemented, or the fire accident determination is made before the time of investigation and evidence collection, determination explanation, approval, etc. of the main evidence", which has a corresponding score of 5 points; in the review rules for the "fire accident determination approval form" document, the review point includes "no record of the fire accident determination approval procedure", which has a corresponding score of 0.2 points.

[0025] Step 1.2: Deconstruct and break down the case file review points, extract and create a structured case file review element library, and establish a mapping relationship between case file review rules, case file review dictionary, case file review elements and judgment logic.

[0026] For example, in the review of the "Fire Accident Determination Approval Form" document, the review point includes "no record of the fire accident determination approval procedure", the review element is "document type = 'Fire Accident Determination Approval Procedure'", the review dictionary is "fire accident determination approval procedure", and the judgment logic is "existence = yes / no"; in the review rules of "enforcement procedure", the review point includes "no record of two or more law enforcement officers jointly conducting a fire accident investigation", the review element is "the number of law enforcement officers in all documents", the review dictionary is "the number of law enforcement officers in a single document", and the judgment logic is "the number of law enforcement officers in a single document is ≥2, and the personnel exist in the list of law enforcement officers".

[0027] Step 1.3: Construct a knowledge graph of case review standards for intelligent review. The knowledge graph includes all case review rules and their corresponding scores, review dictionary, review elements and judgment logic.

[0028] Step 1.4: Synchronize the case file review standard knowledge graph of the intelligent review to the algorithm engineer. Based on the constructed case file review standard knowledge graph, the algorithm engineer uses graph traversal and entity mapping algorithms to convert the structured review elements, review dictionary constraints, and judgment logic into DSL (Dynamic Scripting Language) code logic executable by the rule engine; relying on large model technology, review rules are written to cover the entire intelligent review rule system for case files; at the same time, based on the actual effect of intelligent review feedback from the training set (case files), the knowledge graph information corresponding to the review rules is dynamically adjusted, and the case file review standard knowledge graph is continuously fine-tuned and updated, ultimately forming an iterative intelligent review standard knowledge graph and a directly callable intelligent review engine, such as... Figure 3 As shown.

[0029] Step 2: Subsequent review utilizes an intelligent review engine to conduct intelligent case file review. The intelligent review engine automatically executes different review operations based on the different review points involved in each review checklist. Based on the review results, a preliminary review report is generated, including a structured review report, review score, and suggestions for supplementary evidence (revision comments). For example... Figure 4 As shown.

[0030] Step 2.1: The fire brigade or battalion uploads fire accident investigation files from the client and uses a default review checklist to conduct an intelligent review of the uploaded files. However, experts from the brigade or battalion are allowed to select the necessary review points from the default review checklist and customize a unified list of file review rules.

[0031] Step 2.2: Utilize the intelligent review engine to conduct an intelligent review of the case file, specifically including: Step 2.2.1: Document format and structure analysis.

[0032] The uploaded case files are formatted and converted into readable text. Natural language processing technology is used to identify document types, titles, paragraphs, tables, signatures, and seals to understand the logical structure of the documents. This includes OCR (Optical Character Recognition) recognition, signature recognition, and seal recognition.

[0033] OCR recognition, based on PaddleOCR (Optical Character Recognition) technology, performs full-text recognition of the case file to obtain the text content and corresponding confidence level of each document in the case file.

[0034] Signature recognition uses OpenCV image processing algorithms to extract the handwritten signature area, identifies the presence of the signature through feature matching, and determines the quality of the signature.

[0035] Seal recognition uses OpenCV image processing algorithms for seal detection and comparison; it automatically locates the red circular seal on the document and performs feature matching with a standard seal template to verify the seal type and authenticity.

[0036] Step 2.2.2: Structured information extraction and element matching.

[0037] Based on the case file review element library defined in the selected case file review rule list, structured information is automatically extracted from the parsed case file text to match predefined review elements and dictionaries. For example, for the review of the issuance date in the "Fire Accident Determination Report", the review dictionary is the "issuance date" extracted from the "Fire Accident Determination Report" document.

[0038] Specifically, this includes the document structuring process and the element extraction process.

[0039] During the document structuring process, the results of OCR recognition are analyzed. Based on predefined document title keywords and layout rules, the case file is divided according to document type to form independent document units. The content of each document is then structured, analyzed, and annotated (such as title, body text, list, etc.).

[0040] During the element extraction process, based on natural language processing and regular rules, text elements (such as time, place, people, and events) and image elements (handwritten signatures, seal images, and on-site photos) are extracted and reviewed from the structured document content.

[0041] Based on rule-based automatic review, each review rule is logically compared with the extracted review elements. If the logical comparison of the review rule passes, full marks are awarded; otherwise, all marks for the corresponding review rule are deducted.

[0042] Step 2.2.3: Summarize the review results of all rules, give the review conclusion and modification suggestions for each review rule based on the review results, and calculate the score result of this review in combination with the weight.

[0043] Step 2.2.4: Automatically generate case file summary.

[0044] Understand the content of the documents, summarize the main contents of the case file, and generate a case file summary.

[0045] Step 2.2.5: Automatically find logical contradictions in the statements in the case file.

[0046] By understanding the content of the documents and combining it with the factual evidence in the documents, we can identify logical contradictions that affect the determination of the case file's content.

[0047] Step 2.2.6: Based on the output of Steps 2.2.3 to 2.2.5, automatically generate a preliminary review report. This report includes a summary of the case file content, logical contradictions, review conclusions on the case file review rules, modification suggestions, and scoring results.

[0048] Step 3: Manual Intervention. The marking experts review the generated review report, confirm the automatic review results, and verify the identified issues and approved items. Elements marked as "uncertain to extract" are manually checked and fine-tuned. For issues requiring subjective judgment that cannot be resolved, the review report is supplemented.

[0049] Step 4: After manual verification and confirmation, a case file review report is generated and a scoring result is issued. Case files are handled in a tiered manner based on their quality and scoring results: files with high scores and excellent quality are included in the selection of outstanding files; files with low scores and poor quality are issued a warning of enforcement errors to the case handlers and urged to rectify within a specified period; for files with serious factual errors, an internal error correction mechanism (review) is immediately activated. The case file review report can help identify issues such as whether the procedures used by grassroots fire investigators in fire investigations were legal, whether the investigation time limits were in accordance with regulations, whether the evidence chain for the conclusions was complete, whether the conclusions conflicted with the evidence, and whether the content of standard legal documents was complete. It can also assist fire investigation review agencies in quickly scoring and reviewing the quality of fire investigation files, identifying loopholes and deficiencies in the fire investigation process, facilitating guidance for grassroots fire investigation work, and improving the capabilities of grassroots fire investigators. Simultaneously, the case file review report assesses the attitude and ability of fire investigators towards fire investigation work, improving the internal performance evaluation of fire and rescue teams and the accountability for enforcement errors.

[0050] Example 2

[0051] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the intelligent fire case file review method based on semantic understanding and rule engine.

[0052] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the intelligent fire case file review method based on semantic understanding and a rule engine as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0053] The processor is used to execute all or part of the steps in the intelligent fire case file review method based on semantic understanding and rule engine as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in an electronic device, as well as application-related data.

[0054] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the intelligent fire case file review method based on semantic understanding and rule engine described in the above embodiments.

[0055] Example 3

[0056] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0057] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the intelligent fire case file review method based on semantic understanding and rule engine described in the various embodiments of this application.

[0058] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) app stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the aforementioned intelligent fire case file review method based on semantic understanding and rule engines.

[0059] Example 4

[0060] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned intelligent fire case file review method based on semantic understanding and rule engine.

[0061] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0062] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the present invention.

Claims

1. A method for intelligent review of fire case files based on semantic understanding and rule engine, characterized in that: Includes the following steps: Step 1: During the initial review of fire accident investigation files, the fire and rescue brigade or battalion shall formulate file review standards in accordance with the "Rules for Reviewing Fire Accident Investigation Files" and the file compilation requirements of each province, and form a knowledge graph of file review standards for intelligent review. If this is not the first review, proceed directly to step 2; Step 2: Subsequent review uses an intelligent review engine to conduct intelligent review of the case files; the intelligent review engine automatically performs different review operations based on the different review points involved in each review checklist; based on the review results, a preliminary review report is generated, including a structured review report, review score, and suggestions for supplementing evidence and making revisions; Step 3: Manual intervention; the marking experts review the generated review report, confirm the automatic review results, and verify the identified problems and passed items; the elements marked "uncertain extraction" are manually checked, corrected and fine-tuned. For issues that cannot be handled and require subjective judgment, supplement the review report with additional content; Step 4: After manual verification and confirmation that there are no errors, generate a case file review report.

2. The intelligent fire case file review method based on semantic understanding and rule engine according to claim 1, characterized in that: In step 1, the specific method for forming the knowledge graph of case review standards for intelligent review is as follows: Step 1.1: Manually sort out the content and requirements of case file review, transform the case file review rules stipulated in the articles into review points, assign them corresponding scores, and form a list of case file review rules for intelligent review; Step 1.2: Deconstruct and break down the case file review points, extract and create a structured case file review element library, and establish a mapping relationship between case file review rules, case file review dictionary, case file review elements and judgment logic; Step 1.3: Construct a knowledge graph of case review standards for intelligent review. The knowledge graph includes all case review rules and their corresponding scores, review dictionary, review elements and judgment logic; Step 1.4: Synchronize the case file review standard knowledge graph of intelligent review with the algorithm engineer. Based on the constructed case file review standard knowledge graph, the algorithm engineer uses graph traversal and entity mapping algorithms to convert the structured review elements, review dictionary constraints, and judgment logic into dynamic scripting language code logic that the rule engine can execute. Relying on large model technology, the engineer writes review rules to cover the intelligent review rule system of all review points. At the same time, based on the actual effect of intelligent review feedback from the training set case files, the engineer dynamically adjusts the knowledge graph information corresponding to the review rules, continuously fine-tunes and updates the case file review standard knowledge graph, and finally forms an iterative intelligent review standard knowledge graph and an intelligent review engine that can be directly called.

3. The intelligent fire case file review method based on semantic understanding and rule engine according to claim 1, characterized in that: The specific method for step 2 is as follows: Step 2.1: The fire and rescue brigade or battalion uploads the fire accident investigation file from the client and uses the default review list to conduct intelligent review of the uploaded file; however, experts from the brigade or battalion are allowed to select the required review points from the default review list and customize a unified list of file review rules. Step 2.2: Utilize the intelligent review engine to conduct an intelligent review of the case file; specifically including: Step 2.2.1: Document format and structure analysis; The uploaded case files are formatted and converted into readable text; natural language processing technology is used to identify document types and areas such as titles, paragraphs, tables, seals, and signatures to understand the logical structure of the documents. Step 2.2.2: Structured information extraction and feature matching; Based on the case file review element library defined in the selected case file review rule list, structured information is automatically extracted from the parsed case file text to match predefined review elements and dictionaries; Based on rule-based automatic review, each review rule is logically compared with the extracted review elements; if the logical comparison of the review rule passes, full marks are awarded; if it fails, all marks for the corresponding review rule are deducted. Step 2.2.3: Summarize the review results of all rules, give the review conclusion of each review rule based on the review results, and calculate the score result of this review in combination with the weight; Step 2.2.4: Automatically generate case file summary; Understand the content of the documents, summarize the main contents of the case file, and generate a case file summary; Step 2.2.5: Automatically identify logical contradictions in the statements within the case file; Understand the statements in the interrogation record and, by combining them with the factual basis in the fire accident investigation report, identify any contradictions that contradict the established facts; Step 2.2.6: Based on the output of Steps 2.2.3 to 2.2.5, automatically generate a preliminary review report. This report includes a summary of the case file content, logical contradictions, review conclusions of the case file review rules, suggestions for supplementing evidence, and scoring results.

4. The intelligent fire case file review method based on semantic understanding and rule engine according to claim 3, characterized in that: The identification process in step 2.2.1 includes optical character recognition, signature recognition, and seal recognition; Optical character recognition (OCR) is used to perform full-text recognition of case files based on PaddlePaddle optical character recognition technology, obtaining the text content and corresponding confidence level of each document in the case file. Signature recognition utilizes OpenCV image processing algorithms to extract handwritten signature regions, identifies the presence of a signature through feature matching, and assesses the quality and consistency of the signature. Seal recognition uses OpenCV image processing algorithms for seal detection and comparison; it automatically locates the red circular seal on the document and performs feature matching with a standard seal template to verify the seal type and authenticity.

5. The intelligent fire case file review method based on semantic understanding and rule engine according to claim 4, characterized in that: Step 2.2.2 specifically includes the document structuring process and the element extraction process; During the document structuring process, the results of optical character recognition are analyzed. Based on predefined document title keywords and layout rules, the case files are divided according to document type to form independent document units. The content of each document is then structured, analyzed, and annotated. During the element extraction process, based on natural language processing and regular rules, text elements and image elements are extracted and reviewed from the structured document content. Text elements include time, place, people, reasons, and behaviors; image elements include handwritten signatures, seal images, location indicator images, and on-site photos.

6. An electronic device, characterized in that: The method includes one or more processors and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the intelligent fire case file review method based on semantic understanding and rule engine as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that: The system stores executable instructions that, when executed, cause the processor to perform the intelligent fire case file review method based on semantic understanding and rule engine as described in any one of claims 1-5.

8. A computer program product, characterized in that: Includes a computer program or instructions that, when executed by a processor, implement the intelligent fire case file review method based on semantic understanding and rule engine as described in any one of claims 1-5.

Images