Bridge detection data automatic extraction and database construction system and method

Abstract

The invention provides a bridge detection data automatic extraction and database construction system and method, which are used for solving the problems of low efficiency, poor consistency, difficulty in forming a standardized information base and the like in the existing bridge detection report processing. The system comprises a report acquisition module, a text preprocessing module, a cue word optimization module, an information extraction module, a data structuring module, a data verification module and a database construction module. The method comprises the steps of report acquisition, text preprocessing, cue word generation and optimization, automatic information extraction, data structured processing, data quality control, database construction and the like. Key information is automatically extracted from a bridge detection report through a large language model technology, a standardized database is constructed, the method has the advantages of being high in automation degree, accurate in information extraction, standardized in data, high in system adaptability and the like, and data support is provided for bridge management and maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge engineering inspection data processing, and in particular to a system and method for automatically extracting key information from bridge inspection reports and constructing a standardized database using artificial intelligence technology. Background Art

[0002] By the end of 2024, my country's total number of highway bridges reached 1.0793 million, with a total length of 95.2882 million meters. Small and medium-span bridges accounted for 82.59%, forming a vast bridge management and maintenance system. As bridges age, many are facing performance degradation, placing higher demands on bridge inspection.

[0003] Traditional bridge inspection report management has the following problems: 1. The amount of information is huge but the format is not uniform. Most bridge inspection reports are in paper or electronic form. Preservation, lack of unified format standards; 2. The test and evaluation results are highly subjective and highly dependent on the professional level and experience of the technicians; 3. The effective information density is low, data integration is difficult, and it is not conducive to horizontal and vertical comparative analysis.

[0004] At present, the processing of bridge inspection reports is still mainly done manually, which has shortcomings such as low efficiency, poor consistency, and difficulty in forming a standardized information database, making it difficult to meet the needs of modern bridge management. Summary of the Invention

[0005] In order to solve the problems existing in the existing bridge inspection report processing, the present invention provides a method based on big data. A system and method for automatic extraction of bridge inspection data and database construction based on language model (LLM) is proposed to realize the automated and standardized extraction of key information from bridge inspection reports and construct a structured database.

[0006] To achieve the above object, the present invention provides the following technical solutions: A bridge inspection data automatic extraction and database construction system, comprising: Report acquisition module: used to collect and receive bridge inspection reports; Text preprocessing module: used to perform text recognition and format unification on the test report; Prompt word optimization module: used to generate professional prompt words for bridge inspection reports through iterative testing; Information extraction module: used to extract bridge static and dynamic data from inspection reports based on a large language model and optimized prompt words; Data structuring module: used to convert the extracted unstructured data into structured data in a standard format; Data verification module: used to identify outliers and check data consistency of extracted data; Database construction module: used to store the verified structured data into the standardized bridge inspection database.

[0007] Preferably, the report acquisition module supports multiple formats including paper scanned version and electronic version.

[0008] Preferably, the prompt word optimization module includes: The sample data collection submodule is used to collect and annotate representative bridge inspection report samples; The initial prompt word generation submodule is used to generate the initial prompt word template based on the labeled sample; The prompt word test submodule is used to perform information extraction test on the sample report using the prompt words; The prompt word optimization submodule is used to modify and optimize the prompt words according to the test results; The prompt word library construction submodule is used to classify and organize the optimized prompt words into a professional prompt word library.

[0009] Preferably, the information extraction module supports flexible calling of multiple large language model APIs, including but not limited to OpenAI GPT series, Claude and domestic large models.

[0010] Preferably, the bridge static data includes basic parameters and structural information of the bridge.

[0011] Preferably, the bridge dynamic data includes detection results and defect data.

[0012] Preferably, the data verification module includes an outlier detection submodule, a data integrity check submodule, a data consistency verification submodule and a professional rule verification submodule.

[0013] A method for automatically extracting bridge inspection data and building a database comprises the following steps: (1) Report acquisition steps: Collect bridge inspection reports from multiple sources and conduct preliminary classification and organization; (2) Text preprocessing step: perform OCR recognition on the test report and unify the text format; (3) Prompt word generation step: Optimize professional prompt words through multiple iterative tests to form a prompt word library for bridge inspection reports; (4) Automatic information extraction step: The optimized prompt words and the target report are input into the large language model to automatically extract the static and dynamic data of the bridge; (5) Data structuring step: converting the extracted unstructured data into structured data in a standard format according to a predefined data model; (6) Data quality control step: perform quality control on the extracted data; (7) Database construction step: The verified structured data is stored in the standardized bridge inspection database to form a bridge status database that can be used for analysis and decision-making.

[0014] Preferably, the prompt word generation step includes: a) generating initial prompt words based on sample reports by combining AI automatic extraction with manual annotation; b) continuously optimizing the prompt words through multiple iterative tests to improve the accuracy of information extraction; c) forming a professional prompt word library for different types of bridges and inspection reports.

[0015] Preferably, the data quality control steps include: detecting outliers based on statistical methods and professional knowledge; checking for missing fields; verifying the logical relationship between related fields; supplementing missing or uncertain data through rule inference or re-extraction; and manually reviewing and confirming problematic data.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention automatically extracts key information from bridge inspection reports through large language model technology and constructs a standardized database. It has the advantages of high degree of automation, accurate information extraction, data standardization, and strong system adaptability, providing data support for bridge management and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 Flow chart of the method of the present invention.

[0018] Figure 2 Diagram of the iterative optimization process generated for the prompt word.

[0019] Figure 3 Construct a flow chart for the bridge inspection database. DETAILED DESCRIPTION

[0020] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0021] like Figures 1 to 3 As shown, a bridge inspection data automatic extraction and database construction system of the present invention includes a report acquisition module, a text preprocessing module, a prompt word optimization module, an information extraction module, a data structuring module, a data verification module and a database construction module.

[0022] The report acquisition module supports bridge inspection report formats in multiple formats, including but not limited to PDF, Word, scanned images, etc. For reports in scanned image format, they must first be converted into a processable text format using OCR technology.

[0023] The prompt word optimization module is a key component of this system, specifically including: (1) Collect and annotate sample data: Select representative bridge inspection report samples, have professionals manually annotate them, and determine the key information fields that need to be extracted; (2) Generate initial prompt words: Generate initial prompt word templates through AI assistance based on manually annotated samples; (3) Cue word test and evaluation: Use the initial cue words to conduct information extraction tests on new sample reports and evaluate the extraction effect; (4) Iterative optimization of prompt words: Based on the test results, the prompt words are modified and optimized to enhance their applicability and accuracy; (5) Construct a prompt word library: classify and organize the optimized prompt words to form a professional prompt word library for different types of bridges and inspection reports.

[0024] The information extraction module is based on large language model technology and supports flexible calls to different APIs, including but not limited to the OpenAI GPT series, Claude, and large models developed by major domestic manufacturers. Users can choose the appropriate language model according to their needs. This module is mainly responsible for automatically extracting the following two types of key information from bridge inspection reports: (1) Bridge static data: including basic bridge parameters (bridge name, location, length, width, span combination, design load, construction date, etc.) and structural information (structure type, main beam cross-section, support type, pier type, abutment type, etc.); (2) Bridge dynamic data: including inspection results (inspection date, inspection facility, condition score, condition rating, etc.) and defect data (defect type, location, number, severity, etc.).

[0025] The data structuring module converts the unstructured or semi-structured data output by the large language model into structured data in a standard format. Specifically, it includes: (1) Data parsing: parsing the text information output by the large language model; (2) Data mapping: mapping the parsed information into a predefined data model; (3) Format conversion: Convert data into a format suitable for storage according to database requirements.

[0026] The data verification module performs quality control on the extracted data, mainly including: (1) Outlier detection: identifying and marking data values ​​that may be erroneous or abnormal; (2) Data integrity check: check whether all necessary fields have been extracted; (3) Data consistency verification: ensure that the extracted data is logically consistent; (4) Professional rule verification: Verify whether the data complies with professional rules based on bridge engineering professional knowledge.

[0027] The database construction module builds a standardized bridge inspection database based on the verified structured data, supporting multiple database types such as relational databases or NoSQL databases. The database design follows the following principles: (1) Preserve data originality: save the original extracted data to ensure traceability; (2) Standardized design: adopt standardized database design patterns; (3) Query efficiency optimization: Optimize the database structure to support efficient query; (4) Scalability considerations: reserve space for possible future expansion of data fields.

[0028] The above description is only a preferred embodiment of the present invention and does not constitute any form of interpretation of the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A bridge inspection data automatic extraction and database construction system, characterized by: include: Report acquisition module: used to collect and receive bridge inspection reports; Text preprocessing module: used to perform text recognition and format unification on the test report; Prompt word optimization module: used to generate professional prompt words for bridge inspection reports through iterative testing; Information extraction module: used to extract bridge static and dynamic data from inspection reports based on a large language model and optimized prompt words; Data structuring module: used to convert the extracted unstructured data into structured data in a standard format; Data verification module: used to identify outliers and check data consistency of extracted data; Database construction module: used to store the verified structured data into the standardized bridge inspection database.

2. The system for automatically extracting and building a bridge inspection data database according to claim 1, characterized in that: The report acquisition module supports multiple formats including paper scanned version and electronic version.

3. The system for automatically extracting and building a bridge inspection data database according to claim 1, characterized in that: The prompt word optimization module includes: The sample data collection submodule is used to collect and annotate representative bridge inspection report samples; The initial prompt word generation submodule is used to generate the initial prompt word template based on the labeled sample; The prompt word test submodule is used to perform information extraction test on the sample report using the prompt words; The prompt word optimization submodule is used to modify and optimize the prompt words according to the test results; The prompt word library construction submodule is used to classify and organize the optimized prompt words into a professional prompt word library.

4. The system for automatically extracting bridge inspection data and building a database according to claim 1, characterized in that: The information extraction module supports flexible calls to multiple large language model APIs, including but not limited to the OpenAI GPT series, Claude, and domestic large models.

5. The system for automatically extracting and building a bridge inspection data database according to claim 1, characterized in that: The bridge static data includes basic parameters and structural information of the bridge.

6. The system for automatically extracting and building a bridge inspection data database according to claim 1 or 5, characterized in that: The bridge dynamic data includes detection results and defect data.

7. The system for automatically extracting bridge inspection data and building a database according to claim 1, characterized in that: The data verification module includes an outlier detection submodule, a data integrity check submodule, a data consistency verification submodule and a professional rule verification submodule.

8. A method for automatically extracting bridge inspection data and building a database, characterized by: The following steps are involved: (1) Report acquisition steps: Collect bridge inspection reports from multiple sources and conduct preliminary classification and organization; (2) Text preprocessing step: perform OCR recognition on the test report and unify the text format; (3) Prompt word generation step: Optimize professional prompt words through multiple iterative tests to form a prompt word library for bridge inspection reports; (4) Automatic information extraction step: The optimized prompt words and the target report are input into the large language model to automatically extract the static and dynamic data of the bridge; (5) Data structuring step: converting the extracted unstructured data into structured data in a standard format according to a predefined data model; (6) Data quality control step: perform quality control on the extracted data; (7) Database construction step: The verified structured data is stored in the standardized bridge inspection database to form a bridge status database that can be used for analysis and decision-making.

9. The method for automatically extracting bridge inspection data and building a database according to claim 8, characterized in that: The prompt word generation steps include: a) generating initial prompt words based on sample reports through a combination of AI automatic extraction and manual annotation; b) continuously optimizing the prompt words through multiple iterative tests to improve the accuracy of information extraction; and c) forming a professional prompt word library for different types of bridges and inspection reports.

10. The method for automatically extracting bridge inspection data and building a database according to claim 8, characterized in that: The data quality control steps include: detecting outliers based on statistical methods and professional knowledge; checking for missing fields; verifying the logical relationship between related fields; supplementing missing or uncertain data through rule inference or re-extraction; and manually reviewing and confirming problematic data.

Images