Unstructured data report filling method and system based on large language model

Abstract

The invention discloses an unstructured data report filling method and system based on a large language model, and belongs to the technical field of artificial intelligence, and the method comprises the steps: preprocessing unstructured data, and generating preprocessed unstructured text data; based on the preprocessed unstructured text data, constructing and executing a semantic structured extraction chain; field mapping and report template binding are carried out based on the constructed semantic structured extraction chain, and a field mapping rule is established; based on a field mapping rule, automatically generating and filling a report, and automatically generating the report according to the extracted structured data; and interpretable proofreading and manual auditing are carried out on the automatically generated report. According to the method, implicit information can be accurately extracted, and semantic inference and association are supported; automatic report filling is achieved, and the manual arrangement workload is remarkably reduced; and the output is explainable, and each field can be positioned to the original text, so that the auditing credibility is improved.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, specifically relating to a method and system for filling in unstructured data reports based on a large language model. Background Technology

[0002] In industries such as power, energy, manufacturing, maintenance, and quality supervision, a large amount of unstructured text data is generated daily, including: maintenance records, operation logs, operating condition descriptions, textual descriptions of fault phenomena, safety supervision event records, monthly reports, weekly reports, daily reports, inspection forms, handwritten records, meeting minutes, equipment operating parameters, and trend descriptions. These texts contain important structured data, such as time, values, equipment names, operating status, and alarm levels. However, due to inconsistent description methods, ambiguous descriptions, vague expressions, and non-standardized expressions often occur, causing the following problems: Manually entering reports is extremely time-consuming, especially in data management scenarios involving large-scale maintenance and repair; data accuracy is unstable, and misunderstandings among different personnel lead to inconsistent reporting results; structured information is difficult to obtain automatically, and traditional NLP rule-based methods (regular expressions / keyword matching) cannot handle complex semantics; existing RPA / NLP systems cannot handle semantic relationships across paragraphs and descriptions, and cannot meet the needs of complex real-world scenarios.

[0003] With the development of large language models, although text understanding is possible, the model output is unstable and unstructured, making it unsuitable for direct report input; traditional extraction tools also lack semantic inference capabilities. Therefore, there is an urgent need for a technical solution that combines the deep semantic understanding capabilities of LLM with the stable and structured extraction capabilities of LangChain-extract to achieve enterprise-level intelligent report filling. Summary of the Invention

[0004] This invention aims to provide a method and system for filling in unstructured data reports based on a large language model. The system automatically extracts structured data from any unstructured text and automatically fills it into report templates. It achieves semantic structured extraction across paragraphs and sentences, automatically identifies information such as numerical values, events, time, devices, and status in the text, automatically generates enterprise reports, and fills in structured fields. It provides interpretable and auditable extraction results, significantly improving filling efficiency, accuracy, and consistency.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for filling in unstructured data reports based on a large language model includes the following steps: Step 1: Preprocess the unstructured data to generate preprocessed unstructured text data; Step 2: Based on the preprocessed unstructured text data, construct and execute a semantic structure extraction chain, which includes: The large language model, as the core module of semantic reasoning, is used for context modeling, semantic understanding, and implicit information inference of text; LangChain-extract, as the structured extraction control module, is used to constrain the output structure of the large language model. Step 3: Based on the constructed semantic structured extraction chain, perform field mapping and bind it to the report template, and establish field mapping rules; Step 4: Based on field mapping rules, automatically generate and fill in reports, automatically generating reports based on the extracted structured data; Step 5: Perform interpretability verification and manual review on the automatically generated reports.

[0006] A further improvement of the present invention is that, in step 1, the unstructured data includes: text files, text converted from scanned documents or images by OCR, handwritten records transcribed into text, and system log text; Unstructured data is preprocessed to generate preprocessed unstructured text data, including: Text cleaning and noise removal; sentence segmentation, paragraphing, and semantic boundary delineation; numerical and unit standardization; time expression standardization; and OCR error correction and semantic completion.

[0007] A further improvement of this invention is that, in step 2, the semantic structure extraction chain operates in the following manner: Define the field schema template corresponding to the target report; under schema constraints, guide the large language model to perform semantic matching and field inference on the text; achieve field-controllable JSON structured output; It supports semantic association across sentences and paragraphs; context-based autocomplete for missing fields; and generates corresponding confidence information for the extraction results.

[0008] A further improvement of this invention lies in defining a field schema template corresponding to the target report, including: "Equipment Name"; "Operating parameters, including temperature, current, and vibration;" "Alarm Level"; "Maintenance measures"; "Event Time"; "Status description"; "Trend change".

[0009] A further improvement of this invention is that, in step 3, the constructed semantic structured extraction chain is used as input, and based on the pre-configured field mapping rules, the extracted fields are mapped to the corresponding fields in the target report template. The field mapping rules are established, including: one-to-one or many-to-one mapping relationship between extracted fields and report fields; numerical unit conversion and format conversion rules; merging or splitting strategies for multiple records; and automatic completion or marking rules for missing fields.

[0010] A further improvement of this invention is that, in step 4, a report is automatically generated based on the extracted structured data, including: Write structured data into the corresponding report fields; dynamically expand the table rows based on the number of extracted records; identify abnormal values ​​or data exceeding the threshold; automatically generate report summaries or explanatory text; the generated reports can be exported as Excel, Word, PDF, or written to the enterprise's internal reporting system or database.

[0011] A further improvement of this invention is that, in step 5, the automatically generated reports are subjected to interpretability verification and manual review, including: the automatically generated reports are subjected to interpretability verification and manual review, wherein the verification and review include: providing the corresponding original text source location for each report field; displaying the field inference basis based on the large language model; providing multi-model or rule verification results; and supporting manual revision and feedback for subsequent model optimization.

[0012] An unstructured data reporting system based on a large language model includes: Data preprocessing unit: preprocesses unstructured data to generate preprocessed unstructured text data; Semantic Structured Extraction Chain Unit: Based on preprocessed unstructured text data, a semantic structured extraction chain is constructed and executed, wherein the semantic structured extraction chain includes: The large language model, as the core module of semantic reasoning, is used for context modeling, semantic understanding, and implicit information inference of text; LangChain-extract, as the structured extraction control module, is used to constrain the output structure of the large language model. Field mapping rule establishment unit: Based on the constructed semantic structured extraction chain, field mapping is performed and bound to the report template to establish field mapping rules; Automatic report generation unit: Based on field mapping rules, it automatically generates and fills in reports, and automatically generates reports based on the extracted structured data; Proofreading and manual review unit: Performs interpretability proofreading and manual review on automatically generated reports.

[0013] A further improvement of the present invention is that, in the data preprocessing unit, the unstructured data includes: text files, text converted from scanned documents or images by OCR, handwritten record transcribed text, and system log text; Unstructured data is preprocessed to generate preprocessed unstructured text data, including: Text cleaning and noise removal; sentence segmentation, paragraphing, and semantic boundary delineation; numerical and unit standardization; time expression standardization; and OCR error correction and semantic completion.

[0014] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method for filling in unstructured data reports based on a large language model.

[0015] Compared with the prior art, the present invention has at least the following beneficial technical effects: This invention provides a method and system for filling in unstructured data reports based on a large language model. By introducing a deep semantic understanding mechanism based on a large language model and combining it with LangChain-extract to construct a controllable structured semantic extraction chain, it performs semantic association and field-level extraction across sentences and paragraphs of unstructured text. This effectively solves the problems of semantic ambiguity, information dispersion, and structural instability that traditional rule-based methods struggle to handle. At the same time, by binding field mapping with report templates, it achieves automatic filling of structured data into standardized reports and provides original text tracing and inference basis for each field. Thus, while ensuring extraction stability and auditability, it significantly improves the automation level, data consistency, and filling efficiency of report filling.

[0016] In summary, this invention provides a method for accurately extracting implicit information, supporting semantic inference and association; automatically filling in reports, significantly reducing manual processing workload; and providing interpretable output, with each field locating the original text, thus improving the credibility of audits. This invention also supports multimodal input (text / image / audio), greatly enhancing the system's applicability. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0019] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0020] In the description of this invention, it should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0024] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] Example 1 like Figure 1 As shown, the present invention provides a method for filling in unstructured data reports based on a large language model, comprising the following steps: Step 1: Preprocess the unstructured data to generate preprocessed unstructured text data; Step 2: Based on the preprocessed unstructured text data, construct and execute a semantic structure extraction chain, which includes: The large language model, as the core module of semantic reasoning, is used for context modeling, semantic understanding, and implicit information inference of text; LangChain-extract, as the structured extraction control module, is used to constrain the output structure of the large language model. Step 3: Based on the constructed semantic structured extraction chain, perform field mapping and bind it to the report template, and establish field mapping rules; Step 4: Based on field mapping rules, automatically generate and fill in reports, automatically generating reports based on the extracted structured data; Step 5: Perform interpretability verification and manual review on the automatically generated reports.

[0026] In step 1 of this embodiment, unstructured data includes: text files, text converted from scanned documents or images by OCR, handwritten transcripts, and system log text; Unstructured data is preprocessed to generate preprocessed unstructured text data, including: Text cleaning and noise removal; sentence segmentation, paragraphing, and semantic boundary delineation; numerical and unit standardization; time expression standardization; and OCR error correction and semantic completion.

[0027] In step 2 of this embodiment, the semantic structure extraction chain operates in the following manner: Define the field schema template corresponding to the target report; under schema constraints, guide the large language model to perform semantic matching and field inference on the text; achieve field-controllable JSON structured output; It supports semantic association across sentences and paragraphs; context-based autocomplete for missing fields; and generates corresponding confidence information for the extraction results.

[0028] In this embodiment, a field schema template corresponding to the target report is defined, including: "Equipment Name"; "Operating parameters, including temperature, current, and vibration;" "Alarm Level"; "Maintenance measures"; "Event Time"; "Status description"; "Trend change".

[0029] In step 3 of this embodiment, the constructed semantic structured extraction chain is used as input. Based on the pre-configured field mapping rules, the extracted fields are mapped to the corresponding fields in the target report template. The field mapping rules are established, including: one-to-one or many-to-one mapping relationship between extracted fields and report fields; numerical unit conversion and format conversion rules; merging or splitting strategies for multiple records; and automatic completion or marking rules for missing fields.

[0030] In step 4 of this embodiment, a report is automatically generated based on the extracted structured data, including: Write structured data into the corresponding report fields; dynamically expand the table rows based on the number of extracted records; identify abnormal values ​​or data exceeding the threshold; automatically generate report summaries or explanatory text; the generated reports can be exported as Excel, Word, PDF, or written to the enterprise's internal reporting system or database.

[0031] In step 5 of this embodiment, the automatically generated reports are subjected to interpretability verification and manual review, including: the automatically generated reports are subjected to interpretability verification and manual review, the verification and review include: providing the corresponding original text source location for each report field; displaying the field inference basis based on the large language model; providing multi-model or rule verification results; supporting manual revision and feedback for subsequent model optimization.

[0032] Example 2 This invention provides a method for filling in unstructured data reports based on a large language model, comprising the following steps: Step 1: Input and preprocessing of unstructured data Input formats may include: text file Scanned documents / images (OCR to text) Handwritten content of the form Automatic text of system logs Preprocessing includes: Text cleaning and noise removal Clause and paragraph division Units should be standardized and formatted (e.g., "30℃" should be consistent with "30 degrees"). Time format standardization OCR error correction and semantic completion Step 2: Constructing a semantic structured extraction chain based on LangChain-extract Define the schema template corresponding to the report fields, for example: "Device Name" Operating parameters (temperature, current, vibration, etc.) Alarm Level "Maintenance Measures" "Event Time" "Status Description" "Trend Change" Implemented via LangChain-extract: Field-controlled JSON structure output; automatic association of information across paragraphs; semantic inference and field completion; multi-parameter and multi-device metric extraction; confidence score generation for review.

[0033] Step 3: Field mapping and report template binding Establish field mapping rules: Extract fields and convert them into report fields. Numerical unit conversion Multiple record merging strategies (e.g., daily reports) The rules for completing missing fields (which can be automatically inferred by LLM). The system supports Excel, Word, and PDF templates, custom enterprise spreadsheets, and database reporting systems. Step 4: Automatic Report Generation and Filling The system automatically generates reports based on the extracted structured data: Autofill fields Expand table rows by template Abnormal values ​​are automatically highlighted in red. Automatically generate summaries (such as daily report summaries). Can be exported or uploaded to enterprise systems Step 5: Explainability proofreading and manual review To enhance audit credibility, the system provides: the original text location for each field; the rationale behind model inferences; multi-model validation results; and a manual revision interface (allowing for reverse model optimization). Example 3 Maintenance records are automatically generated into structured archives. Input handwritten content (OCR to text): "The bearing of the No. 2 feedwater pump was replaced and found to be severely worn. It is estimated that it can continue to operate for another 500 hours. The operating status has returned to normal." System entry: Repair item: Bearing replacement; Cause of failure: Severe wear; Expected service life: 500 hours; Current status: Normal.

[0034] Taking the equipment maintenance management scenario of a thermal power plant as an example, a standardized "Daily Equipment Maintenance Report" needs to be generated daily to record the maintenance status of key equipment in the unit. The report fields include equipment name, maintenance content, abnormal phenomena, cause analysis, handling measures, key operating parameters, estimated uptime, and current operating status. The traditional method requires manual reading of maintenance records and filling in each item, which is inefficient and inconsistent.

[0035] (a) Input of raw unstructured data The unstructured data received by the system comes from handwritten maintenance records filled out by maintenance personnel on-site. After scanning, the text content is obtained through OCR recognition: "At 8:30 on June 12, 2025, maintenance was carried out on feedwater pump A of Unit 2. During operation, it was found that the bearing temperature rose to 92 degrees Celsius and the vibration increased significantly. After disassembly and inspection, it was found that the bearing was severely worn. A new bearing was replaced and lubricating oil was added. At 10:10, the test run was normal, the temperature returned to 65 degrees Celsius, and the vibration returned to the normal range. It is expected to run continuously for 500 hours." The text contains time information, device information, anomaly description, numerical parameters, processing actions, and operating status. Its semantics are distributed across multiple sentences and paragraphs, making it impossible to extract completely using simple rules.

[0036] (II) Step 1: Unstructured data preprocessing The system first performs preprocessing operations on the above text, including: OCR error correction (e.g., uniformly recognizing "92 degrees" as "92℃"); Time standardization ("June 12, 2025, 8:30" is converted to a unified timestamp format); Unit normalization ("degree" and "℃" are unified as "℃"); Sentence segmentation and semantic division generate structured semantic units.

[0037] The preprocessed text is then fed into the subsequent semantic extraction chain.

[0038] (III) Step 2: Semantic Structure Extraction Based on Large Language Model and LangChain-extract The system pre-defines a structured schema template for the "Equipment Maintenance Daily Report," for example: Equipment Name Maintenance time Abnormal phenomena Key operating parameters (temperature, vibration) Cause of the fault Maintenance measures Trial operation results Current running status Expected uptime A semantic extraction chain built on LangChain-extract calls a large language model to perform deep semantic understanding on the preprocessed text and outputs structured JSON data strictly according to schema constraints. An example of the extraction results is shown below: { Equipment Name: Unit 2 Feedwater Pump A Maintenance Time: 2025-06-12 08:30 Abnormal phenomenon: "Bearing temperature rises, vibration increases". "Key operating parameters": { "Bearing Temperature Abnormality": "92℃", "Bearing Temperature Recovery": "65℃", "Vibration Status": "Returned to Normal" }, Cause of failure: Severe bearing wear Repair measures: Replace the bearing and add lubricating oil. Trial run results: "Trial run was normal". Current running status: "Normal" "Estimated uptime": "500 hours", "Draw confidence level": 0.94 } Among them, fields such as "estimated run time" and "current running status" do not appear in a fixed format, but are automatically generated by combining the semantic inference capabilities of the large language model with the context.

[0039] (iv) Step 3: Field mapping and report template binding The system maps the above structured data to the fields of the enterprise's standard "Equipment Maintenance Daily Report" template according to the preset field mapping rules, and then executes the following: Field name mapping (e.g., "Equipment Name" → "Equipment Under Maintenance") Numerical field validation and unit retention; A single maintenance record automatically generates a single line of daily report data.

[0040] (v) Step 4: Automatic report generation and filling The system automatically writes structured data into the daily report template and generates complete report content, including: Automatically fill in fields such as maintenance time, equipment name, and maintenance content; Mark "bearing temperature abnormal value 92℃" as an abnormal parameter; Automatically generate daily report summary statements, such as: "The abnormality of bearing A of feedwater pump of Unit 2 has been dealt with and the equipment operation status has returned to normal." The generated reports can be directly exported to Excel or uploaded to the enterprise maintenance management system.

[0041] (vi) Step 5: Explainability proofreading and manual review In the manual review interface, the system provides the following for each field: The corresponding original text source location; Explanation of the inference basis for generating this field using the large language model; Extract confidence level indicators.

[0042] Reviewers can revise individual fields, and the revision results will be used as feedback data for continuous optimization of the subsequent model and extraction chain.

[0043] Example 4 like Figure 2 As shown, the present invention provides an unstructured data report filling system based on a large language model, comprising: Data preprocessing unit: preprocesses unstructured data to generate preprocessed unstructured text data; Semantic Structured Extraction Chain Unit: Based on preprocessed unstructured text data, a semantic structured extraction chain is constructed and executed, wherein the semantic structured extraction chain includes: The large language model, as the core module of semantic reasoning, is used for context modeling, semantic understanding, and implicit information inference of text; LangChain-extract, as the structured extraction control module, is used to constrain the output structure of the large language model. Field mapping rule establishment unit: Based on the constructed semantic structured extraction chain, field mapping is performed and bound to the report template to establish field mapping rules; Automatic report generation unit: Based on field mapping rules, it automatically generates and fills in reports, and automatically generates reports based on the extracted structured data; Proofreading and manual review unit: Performs interpretability proofreading and manual review on automatically generated reports.

[0044] In the data preprocessing unit of this embodiment, unstructured data includes: text files, text converted from scanned documents or images by OCR, handwritten records transcribed into text, and system log text; Unstructured data is preprocessed to generate preprocessed unstructured text data, including: Text cleaning and noise removal; sentence segmentation, paragraphing, and semantic boundary delineation; numerical and unit standardization; time expression standardization; and OCR error correction and semantic completion.

[0045] Example 5 The present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for filling in unstructured data reports based on a large language model.

[0046] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] This application is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0048] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0049] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0050] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0051] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for filling in unstructured data reports based on a large language model, characterized in that, Includes the following steps: Step 1: Preprocess the unstructured data to generate preprocessed unstructured text data; Step 2: Based on the preprocessed unstructured text data, construct and execute a semantic structure extraction chain, which includes: Large language models serve as the core module for semantic reasoning, used for contextual modeling, semantic understanding, and implicit information inference of text. LangChain-extract serves as the structured extraction control module, used to constrain the output structure of large language models. Step 3: Based on the constructed semantic structured extraction chain, perform field mapping and bind it to the report template, and establish field mapping rules; Step 4: Based on field mapping rules, automatically generate and fill in reports, automatically generating reports based on the extracted structured data; Step 5: Perform interpretability verification and manual review on the automatically generated reports.

2. The method for filling in unstructured data reports based on a large language model according to claim 1, characterized in that, In step 1, unstructured data includes: text files, scanned documents or images converted by OCR, handwritten records transcribed into text, and system log text; Unstructured data is preprocessed to generate preprocessed unstructured text data, including: Text cleaning and noise removal; sentence segmentation, paragraphing, and semantic boundary delineation; numerical and unit standardization; time expression standardization; and OCR error correction and semantic completion.

3. The method for filling in unstructured data reports based on a large language model according to claim 1, characterized in that, In step 2, the semantic structure extraction chain works in the following way: Define the field schema template corresponding to the target report; under schema constraints, guide the large language model to perform semantic matching and field inference on the text; achieve field-controllable JSON structured output; Supports semantic association across sentences and paragraphs; Context-based autocomplete for missing fields; And generate corresponding confidence information for the extraction results.

4. The method for filling in unstructured data reports based on a large language model according to claim 3, characterized in that, Define the field schema template corresponding to the target report, including: "Device Name"; Operating parameters, including temperature, current, and vibration; "Alarm Level"; "Maintenance measures"; "Event Time"; "Status description"; "Trend change".

5. The method for filling in unstructured data reports based on a large language model according to claim 1, characterized in that, In step 3, the constructed semantic structured extraction chain is used as input. Based on the pre-configured field mapping rules, the extracted fields are mapped to the corresponding fields in the target report template. The field mapping rules are established, including: one-to-one or many-to-one mapping relationship between extracted fields and report fields; rules for numerical unit conversion and format conversion; strategies for merging or splitting multiple records; and rules for automatic completion or marking of missing fields.

6. The method for filling in unstructured data reports based on a large language model according to claim 1, characterized in that, In step 4, reports are automatically generated based on the extracted structured data, including: Write structured data into the corresponding report fields; dynamically expand the table rows based on the number of extracted records; identify abnormal values ​​or data exceeding the threshold; automatically generate report summaries or explanatory text; the generated reports can be exported as Excel, Word, PDF, or written to the enterprise's internal reporting system or database.

7. The method for filling in unstructured data reports based on a large language model according to claim 1, characterized in that, In step 5, the automatically generated reports undergo interpretability verification and manual review, including: providing the original text source location for each report field; displaying the field inference basis based on the large language model; providing multi-model or rule verification results; and supporting manual revision and feedback for subsequent model optimization.

8. A system for filling out unstructured data reports based on a large language model, characterized in that, include: Data preprocessing unit: preprocesses unstructured data to generate preprocessed unstructured text data; Semantic Structured Extraction Chain Unit: Based on preprocessed unstructured text data, a semantic structured extraction chain is constructed and executed, wherein the semantic structured extraction chain includes: The large language model, as the core module of semantic reasoning, is used for context modeling, semantic understanding, and implicit information inference of text; LangChain-extract, as the structured extraction control module, is used to constrain the output structure of the large language model. Field mapping rule establishment unit: Based on the constructed semantic structured extraction chain, field mapping is performed and bound to the report template to establish field mapping rules; Automatic report generation unit: Based on field mapping rules, it automatically generates and fills in reports, and automatically generates reports based on the extracted structured data; Proofreading and manual review unit: Performs interpretability proofreading and manual review on automatically generated reports.

9. The unstructured data reporting system based on a large language model according to claim 8, characterized in that, In the data preprocessing unit, unstructured data includes: text files, scanned documents or images converted by OCR to text, handwritten records transcribed to text, and system log text; Unstructured data is preprocessed to generate preprocessed unstructured text data, including: Text cleaning and noise removal; sentence segmentation, paragraphing, and semantic boundary delineation; numerical and unit standardization; time expression standardization; and OCR error correction and semantic completion.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the unstructured data report filling method based on a large language model according to any one of claims 1-7.

Images