Case element extraction method based on natural language processing technology
Through the case element extraction method based on natural language processing technology, the problem of difficult factor association relationships in the existing technology is solved, the intelligent identification and relationship establishment of case elements are realized, the efficiency of case investigation is improved, and in-depth analysis support is provided for the public security. .
Patent Information
- Application Number
- CN202411078203.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-12-03
AI Technical Summary
The existing technology is difficult to explore the correlation relationship between elements in case factor data extraction, resulting in low accuracy of factor extraction, easy to omit or error extraction, unable to adapt to complex natural language text case descriptions, and unable to meet the public security needs for in-depth analysis and mining of cases. .
Using natural language processing technology methods, including data collection and preprocessing, feature selection, construction of configurable annotation tools, case element annotation, identification and extraction model selection and construction, model training, evaluation and testing, deployment and monitoring, through Deep learning algorithms build case element recognition and extraction models to realize intelligent identification of factor entities and relationships.
It has improved the intelligent identification and relationship establishment of case elements, improved the investigators' ability to quickly analyze the factors involved in the case, obtained valuable relationships, improved the efficiency of case investigation, and provided decision-making support for the public security for in-depth exploration and risk warning.
Smart Images

Figure CN119067106A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of entity recognition and keyword extraction, case element recognition and extraction model selection and construction, post-processing and integration, and specifically, to a case element extraction method based on natural language processing technology. Background Art
[0002] In recent years, non-contact property crimes have been on the rise. In such cases, analysis is usually carried out around the suspects, crime locations, institutions, financial information, financial relationships, virtual identities, apps, etc. By extracting case element data and element association rule data, we support case analysis and judgment and improve the accuracy of case mining.
[0003] At present, the main method for extracting case element data is still focused on means such as keyword text segmentation. This method is difficult to mine the correlation between elements and cannot adapt to the case description of complex natural language text. The accuracy of element extraction is low, and it is easy to miss elements in the case or extract elements incorrectly.
[0004] Therefore, an effective and accurate method and tool for identifying and extracting case information elements and relationships is needed to meet the public security needs for in-depth analysis and mining of cases. Summary of the invention
[0005] In view of the problems in the related technology, the present invention proposes a case element extraction method based on natural language processing technology to overcome the above-mentioned technical problems existing in the existing related technology.
[0006] To this end, the specific technical solution adopted by the present invention is as follows: A case element extraction method based on natural language processing technology includes the following steps: Step S1: case data collection and preprocessing; Step S2: case element feature selection; Step S3: building a configurable annotation tool; Step S4: case element labeling; Step S5: case element identification and extraction model selection and construction; Step S6: model training; Step S7: model evaluation and testing; Step S8: Model deployment and monitoring.
[0007] As a further solution of the present invention, a large amount of case-related data is collected in step S1 and preprocessed. The processing content includes data cleaning, feature extraction and data labeling, etc. Data cleaning and feature processing are performed through regular expressions (Regex).
[0008] As a further solution of the present invention, the case element feature selection in step S2 refers to selecting element feature variables related to the case from the original data. The main case element features include information about the person involved in the case, the time involved in the case, the location of the case, the organization involved in the case, the fund information involved in the case, the transfer relationship, the virtual identity information, the personnel relationship, the item information, the case characteristics, the modus operandi, and other information.
[0009] As a further solution of the present invention, in step S3, a labeling tool is established to establish common case element entities and relationship labels, provide case text selection labeling, review, and quality inspection functions, and provide manual labeling, review, and quality inspection. A real-time visual management system is established to improve human-computer collaboration and labeling efficiency.
[0010] As a further solution of the present invention, the case element labeling process in step S4: screening and reviewing case information -> pre-labeling that meets the conditions -> manual labeling -> multiple quality checks -> re-labeling of unqualified ones -> determining the labeled data; Named element entity annotation: Use annotation methods such as key point annotation and semantic segmentation annotation to annotate and classify element entities in case text data; Relationship annotation: Annotate the relationships of case information. The main relationships include the relationship between people involved in the case, the relationship between people and virtual identities, the relationship between fund transfers, the relationship between people and items, the relationship between people and institutions, etc. Semantic annotation: Semantic annotation confirms the meaning of concepts such as case category, crime characteristics, and crime methods that appear in the case text and labels them.
[0011] As a further solution of the present invention, in step S5, the characteristics of the case elements are taken into consideration, an algorithm based on natural language processing (NLP) deep learning is selected, and a case element recognition and extraction model is constructed.
[0012] As a further solution of the present invention, in step S6, a large amount of data is input for learning and training, so that the model can identify and classify the input data. The model patrols using supervised learning and unsupervised learning methods; Supervised learning model training: Use the labeled case data and corresponding output results for learning and training, gradually adjust the model parameters, learn the element entities in the data and the association between the data, so as to accurately identify and classify new data; Unsupervised learning model training: the model learns and discovers autonomously based on the inherent structure and characteristics of the input data. The recognition results are manually checked, corrected, and the model optimized.
[0013] As a further solution of the present invention, in step S7, different test data are used to verify and test the trained model to evaluate the recognition ability and accuracy of the model. For areas where the model performs poorly, the model is fine-tuned and optimized.
[0014] As a further solution of the present invention, in step S8, the trained model and tools are deployed to the production environment to ensure that newly collected case data can be processed in real time. A monitoring tool is established to monitor the operation and performance of the model in real time to ensure the stability and reliability of the operation; Through this case element extraction method, it is possible to intelligently identify and extract elements of case data and establish relationships, providing decision-making support for public security personnel in deep case mining and risk warning.
[0015] Beneficial Effects of the Invention The present invention is based on natural language processing deep learning technology, provides a method and tool based on intelligent case element entity recognition and relationship recognition, and adaptively and optimally adjusts system parameters and algorithms. Through this invention, investigators can quickly analyze the elements involved in the case, obtain valuable correlation relationships, and improve the efficiency of case detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 A flowchart of a case element extraction method based on natural language processing technology of the present invention; DETAILED DESCRIPTION
[0017] According to an embodiment of the present invention, a case element extraction method based on natural language processing technology is provided.
[0018] A case element extraction method based on natural language processing technology according to an embodiment of the present invention comprises the following steps: Step S1: case data collection and preprocessing; Step S2: case element feature selection; Step S3: building a configurable annotation tool; Step S4: case element labeling; Step S5: case element identification and extraction model selection and construction; Step S6: model training; Step S7: model evaluation and testing; Step S8: Model deployment and monitoring.
[0019] In an optional embodiment, a large amount of case-related data is collected in step S1 and preprocessed. The processing includes data cleaning, feature extraction, data labeling, etc. Data cleaning and feature processing are performed through regular expressions (Regex).
[0020] In an optional embodiment, the case element feature selection in step S2 refers to selecting element feature variables related to the case from the original data. The main case element features include information about the person involved in the case, the time involved in the case, the location of the case, the organization involved in the case, the fund information involved in the case, the transfer relationship, the virtual identity information, the personnel relationship, the item information, the case characteristics, the modus operandi, and other information.
[0021] In an optional embodiment, in step S3, a labeling tool is established to establish common case element entities and relationship labels, provide case text selection labeling, review, and quality inspection functions, and provide manual labeling, review, and quality inspection. A real-time visual management system improves human-computer collaboration and labeling efficiency.
[0022] In an optional embodiment, the case element labeling process in step S4: screening and reviewing case information -> pre-labeling that meets the conditions -> manual labeling -> multiple quality checks -> re-labeling of unqualified ones -> determining the labeled data; Named element entity annotation: Use annotation methods such as key point annotation and semantic segmentation annotation to annotate and classify element entities in case text data; Relationship annotation: Annotate the relationships of case information. The main relationships include the relationship between people involved in the case, the relationship between people and virtual identities, the relationship between fund transfers, the relationship between people and items, the relationship between people and institutions, etc. Semantic annotation: Semantic annotation confirms the meaning of concepts such as case category, crime characteristics, and crime methods that appear in the case text and labels them.
[0023] In an optional embodiment, in step S5, the characteristics of the case elements are taken into consideration, an algorithm based on natural language processing (NLP) deep learning is selected, and a case element recognition and extraction model is constructed.
[0024] In an optional embodiment, in step S6, a large amount of data is input for learning and training, so that the model can identify and classify the input data. The model patrols using supervised learning and unsupervised learning methods; Supervised learning model training: Use the labeled case data and corresponding output results for learning and training, gradually adjust the model parameters, learn the element entities in the data and the association between the data, so as to accurately identify and classify new data; Unsupervised learning model training: the model learns and discovers autonomously based on the inherent structure and characteristics of the input data. The recognition results are manually checked, corrected, and the model optimized.
[0025] In an optional embodiment, in step S7, different test data are used to verify and test the trained model to evaluate the recognition ability and accuracy of the model. For areas where the model performs poorly, the model is fine-tuned and optimized.
[0026] In an optional embodiment, in step S8, the trained model and tools are deployed to the production environment to ensure that newly collected case data can be processed in real time. A monitoring tool is established to monitor the operation and performance of the model in real time to ensure the stability and reliability of the operation; Through this case element extraction method, it is possible to intelligently identify and extract elements of case data and establish relationships, providing decision-making support for public security personnel in deep case mining and risk warning.
[0027] The present invention can be applied to: The present invention is based on natural language processing deep learning technology, provides a method and tool based on intelligent case element entity recognition and relationship recognition, and adaptively and optimally adjusts system parameters and algorithms. Through this invention, investigators can quickly analyze the elements involved in the case, obtain valuable correlation relationships, and improve the efficiency of case detection.
[0028] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A case element extraction method based on natural language processing technology, It is characterized in that The following steps are involved: Step S1: case data collection and preprocessing; Step S2: case element feature selection; Step S3: building a configurable annotation tool; Step S4: case element labeling; Step S5: case element identification and extraction model selection and construction; Step S6: model training; Step S7: model evaluation and testing; Step S8: Model deployment and monitoring.
2. According to the method for extracting case elements based on natural language processing technology in claim 1, Features: In step S1, a large amount of case-related data is collected and preprocessed. The processing content includes data cleaning, feature extraction, and data labeling. Data cleaning and feature processing are performed through regular expressions (Regex).
3. According to the case element extraction method based on natural language processing technology as described in claim 1, Features: The case element feature selection in step S2 refers to selecting case-related element feature variables from the original data. The main case element features include information about the person involved in the case, the time involved in the case, the location of the case, the organization involved in the case, the fund information involved in the case, the transfer relationship, the virtual identity information, the personnel relationship, the item information, the case characteristics, the modus operandi, and other information.
4. According to the method for extracting case elements based on natural language processing technology as described in claim 1, Features: In step S3, a labeling tool is established to establish common case element entities and relationship labels, provide case text selection labeling, review, and quality inspection functions, and manual labeling, review, and quality inspection. A real-time visual management system is used to improve human-machine collaboration and labeling efficiency.
5. According to the method for extracting case elements based on natural language processing technology in claim 1, Features: The case element labeling process in step S4: screening and reviewing case information -> pre-labeling that meets the conditions -> manual labeling -> multiple quality checks -> re-labeling of unqualified ones -> confirming the labeled data; Named element entity annotation: Use annotation methods such as key point annotation and semantic segmentation annotation to annotate and classify element entities in case text data; Relationship annotation: Annotate the relationships of case information. The main relationships include the relationship between people involved in the case, the relationship between people and virtual identities, the relationship between fund transfers, the relationship between people and items, the relationship between people and institutions, etc. Semantic annotation: Semantic annotation confirms the meaning of concepts such as case category, crime characteristics, and crime methods that appear in the case text and labels them.
6. The case element extraction method based on natural language processing technology according to claim 1, Features: In step S5, the characteristics of the case elements are taken into consideration, an algorithm based on natural language processing (NLP) deep learning is selected, and a case element recognition and extraction model is constructed.
7. The case element extraction method based on natural language processing technology according to claim 1, Features: In step S6, a large amount of data is input for learning and training, so that the model can identify and classify the input data. The model patrols using supervised learning and unsupervised learning methods; Supervised learning model training: Use the labeled case data and corresponding output results for learning and training, gradually adjust the model parameters, learn the element entities in the data and the association between the data, so as to accurately identify and classify new data; Unsupervised learning model training, the model autonomously learns and discovers based on the intrinsic structure and characteristics of the input data. The recognition results are manually checked, corrected and the model optimized.
8. The case element extraction method based on natural language processing technology according to claim 1, Features: In step S7, different test data are used to verify and test the trained model to evaluate the recognition ability and accuracy of the model. For areas where the model performs poorly, the model is fine-tuned and optimized.
9. The case element extraction method based on natural language processing technology according to claim 1, Features: In step S8, the trained model and tools are deployed to the production environment to ensure that newly collected case data can be processed in real time. Monitoring tools are established to monitor the operation and performance of the model in real time to ensure the stability and reliability of the operation; Through this case element extraction method, it is possible to intelligently identify and extract elements of case data and establish relationships, providing decision-making support for public security personnel in deep case mining and risk warning.
Citation Information
Cited By
Decision document element extraction method based on large model, storage medium and equipment
CN121787423A
