A system and method for improving the quality of investigation final report based on digital model
Through a digital model-based method, the investigation final report is automatically cut and feature analysis to generate intelligent layout reports, solving the problems of low efficiency and poor quality of report writing in the existing technology, and achieving efficient and reliable report generation.
Patent Information
- Application Number
- CN202410762604.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-13
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-06-13
AI Technical Summary
In the prior art, the efficiency of writing investigation final reports is low, poor quality, irregular format and repeated reviews, resulting in limited efficiency and quality of case handlers in the process of writing reports.
Using a digital model-based method, the paragraph automatic cutting and entity information extraction of the historical investigation end report is performed, and feature extraction and association combination analysis is performed based on the structured data of the current case to generate an intelligent typed investigation end report.
It greatly improves the efficiency of report writing of case handlers, reduces the steps of data inquiry, improves the reliability and quality of reports, and overall improves the efficiency and quality of report writing.
Smart Images

Figure CN118607482B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data analysis, and in particular to a system and method for improving the quality of investigation final reports based on a digital model. Background Art
[0002] Writing an investigation termination report is a necessary step for administrative departments to handle cases. In the current technical environment, the commonly used method is to use the system to automatically generate an investigation termination report template. This method is for case handlers to write the report content online. However, this model has major drawbacks, including low writing efficiency, poor report quality, irregular writing format, and high review repetition. Summary of the invention
[0003] The purpose of the present invention is to provide a system and method for improving the quality of investigation final report based on a digital model to solve the problems raised in the above-mentioned background technology.
[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0005] A method for improving the quality of investigation final report based on a digital model, the method comprising the following steps:
[0006] S100, automatically segment the historical investigation final report into paragraphs, send the text segmentation results to the model for training after manual review, extract report entity information from the investigation final report text, and integrate the extracted information to generate paragraph text;
[0007] S200, extracting features from the structured data of the current case according to the paragraph text data, and obtaining feature group data of the paragraph text of the investigation finalization report corresponding to the current case;
[0008] S300, performing association combination analysis based on the feature group data of the paragraph text of the current case combined with historical training data, obtaining relevant evidence group data, discretion group data and handling opinion group data of the current case; and performing a comprehensive combination evaluation of the content text of the current case combined with the obtained data;
[0009] S400: intelligently typeset the acquired data in combination with the paragraph text of the case according to the evaluation result and output the investigation completion report of the current case.
[0010] The S100 automatically segments the historical investigation final report into paragraphs, sends the text segmentation results to the model for training after manual review, extracts report entity information from the investigation final report text, and integrates the extracted information to generate paragraph text. The specific steps are as follows:
[0011] S101, constructing a historical data sample by retrieving the historical investigation final report data in the database; segmenting the historical sample data into paragraphs by using the automatic segmentation process; the automatic segmentation process is to use the paragraph segmentation model to realize the automatic segmentation of the text into paragraphs; manually review the segmented sample segmentation data, and after passing the review, input the data into the training model for training; the training model is to call the text training model to realize data recognition and classification of the segmentation data;
[0012] S102, extracting entity information data of the investigation finalization report in combination with the sample segmentation data, and integrating the retrieved entity information to generate corresponding investigation report paragraph texts; the entity information includes people, places, time events, etc.
[0013] The specific steps of performing feature extraction on the structured data of the current case according to the paragraph text data and obtaining the feature group data of the paragraph text of the investigation finalization report corresponding to the current case in S200 are as follows:
[0014] S201, extracting structured data of the current case based on the paragraph text data of the historical sample data; the structured data corresponds to the entity data in the historical sample data; and performing feature processing on the extracted structured data, constructing a corresponding feature matrix for each structured data Where n is the number of structured data type labels, is the evaluation characteristic value of the corresponding feature point after the characterization processing of each type of structured data, and m is the number of corresponding feature points after the characterization processing of each type of structured data; by comparing the sample paragraph data, the feature difference analysis of the current case structured data is performed, and the analysis formula is: where dov n is the difference between the sample data corresponding to each type of structured data in the current case; C n is the sample data feature matrix corresponding to each type of structured data of the current case; and are the feature vectors generated by the feature matrix of the current case and the feature matrix of the sample data respectively; J(A n ) and J(C n ) are the variance values of the elements in the current case feature matrix and the sample data feature matrix respectively; m(A n ) and m(C n ) are the number of elements in the current case feature matrix and the sample data feature matrix respectively;
[0015] S202, manually set the judgment threshold dov s , perform feature screening on various types of structured data of the current case. If dov n ≥dovs , then the corresponding type of structured data is judged to be the difference feature data of the current case; if dov n <dov s , then the corresponding type of structured data is judged to be the conventional feature data of the current case; according to the judgment result, the feature matrix corresponding to the structured data of the difference feature data of the current case is combined, and the combination result is Among them A D is the difference feature matrix after combined processing; is the feature matrix corresponding to the structured data corresponding to the difference feature data; c is the type of difference feature data in the corresponding n types of structured data; c is the type data label of the difference feature data.
[0016] The S300 performs an associated combination analysis based on the feature group data of the paragraph text of the current case in combination with the historical training data to obtain the relevant evidence group data, discretion group data and handling opinion group data of the current case; the specific steps of performing a comprehensive combination evaluation on the content text of the current case in combination with the obtained data are as follows:
[0017] S301: Perform similarity matching between the conventional feature data in the current case and the historical sample data in the database, and filter the historical sample data according to the matching results. The matching analysis formula is: Among them, Sim i is the matching degree of the conventional feature data and sample data in the current case; Calculates vector dot product; The feature vector generated by the feature matrix corresponding to each conventional feature data in the sample data; is the feature vector generated by the feature matrix corresponding to each conventional feature data in the current case, where i is the number label of the conventional feature data type, i=nc; according to the analysis results, the sample data corresponding to the maximum matching degree between each type of conventional feature data and the sample data is obtained in descending order, and the sample data is extracted to construct a comparison sample set [B1, B2, B3...B i ]; among them B i The sample data corresponding to the maximum matching degree of each type of conventional feature data collected;
[0018] S302, by analyzing and comparing the difference feature data of each sample data in the sample set and the current case, the relevant evidence group data, discretion group data and processing opinion group data corresponding to each type of structured data are calculated, and the calculation formula is: in The degree of fit between each difference feature data and each sample data in the comparison sample set; The feature vector generated by the feature matrix corresponding to each difference feature data of the current case; is the eigenvector generated by the eigenmatrix corresponding to each difference feature data of each sample data in the comparison sample set; α and β are the horizontal angles of the eigenvector generated by the eigenmatrix corresponding to each difference feature data of each sample data in the current case and the comparison sample set; according to the calculation results of the degree of fit between each difference feature data and each sample data in the comparison sample set, select the sample data in the comparison sample set corresponding to the maximum value of the calculated degree of fit of each difference feature data, collect the corresponding relevant evidence data, discretionary data and processing opinion data for each sample data, and coordinate the collected data of each sample data to obtain comprehensive relevant evidence group data, discretionary group data and processing opinion group data; because when calculating the degree of fit, each difference feature data of the current case needs to be calculated with each sample data in the comparison sample set, so the degree of fit calculation of each difference feature data is multiple, so the maximum value is selected to extract the relevant data as the closest and most reliable data;
[0019] S303: Perform reliability analysis on the current case based on the obtained comprehensive relevant evidence group data, discretion group data, and processing opinion group data corresponding to the sample data. The calculation formula is: Where e is the number of sample data in the comparative sample set that collects relevant evidence, discretion, and processing opinion data; The feature vector generated for the feature matrix of sample data in the comparison sample set corresponding to the collection of relevant evidence, discretion and processing opinion data.
[0020] The specific steps of S400 combining the acquired data with the paragraph text of the case for intelligent typesetting and outputting the investigation finalization report of the current case according to the evaluation result are as follows:
[0021] S401, intelligently typeset the relevant evidence, discretion and handling opinion data obtained by matching the current case with the historical sample data and the paragraph text of the current case to generate an investigation completion report corresponding to the current case;
[0022] S402: Output the investigation completion report corresponding to the current case and its corresponding reliability analysis results and prompt the reviewer to confirm.
[0023] A system for improving the quality of investigation final reports based on a digital model, the system comprising a model processing module, a feature data processing module, a report comprehensive analysis module and a report output module;
[0024] The model processing module automatically segments the historical investigation termination report into paragraphs, and after manual review, the text segmentation results are sent to the model for training, the report entity information is extracted from the investigation termination report text, and the extracted information is integrated to generate paragraph text; the feature data processing module performs feature extraction on the current case structured data based on the paragraph text data, and obtains the feature group data of the investigation termination report paragraph text corresponding to the current case; the report comprehensive analysis module performs association combination analysis based on the feature group data of the current case paragraph text in combination with the historical training data, and obtains the relevant evidence group data, discretion group data and handling opinion group data of the current case; the content text of the current case is comprehensively evaluated in combination with the acquired data; the report output module combines the acquired data with the case paragraph text for intelligent typeset according to the evaluation results and outputs the investigation termination report of the current case.
[0025] The model processing module includes a historical data cutting unit and an entity data extraction unit; the historical data cutting unit constructs a historical data sample by retrieving the historical investigation termination report data from the database; the historical sample data is segmented by using automatic paragraph segmentation processing; the segmented sample cutting data is manually reviewed, and after the review, the data is input into the training model for training; the entity data extraction unit extracts the entity information data of the investigation termination report in combination with the sample cutting data, and integrates the retrieved entity information to generate corresponding investigation report paragraph texts.
[0026] The feature data processing module includes a data difference analysis unit and a difference data judgment unit; the data difference analysis unit extracts structured data of the current case based on the paragraph text data of the historical sample data; and performs feature processing on the extracted structured data, and constructs a corresponding feature matrix for each structured data; performs feature difference analysis on the structured data of the current case by comparing the sample paragraph data; the difference data judgment unit performs feature screening on various types of structured data of the current case by manually setting a judgment threshold; and according to the judgment result, combines and processes the feature matrices corresponding to the structured data that matches the difference feature data of the current case.
[0027] The report comprehensive analysis module includes a conventional data matching analysis unit, a relevant data extraction analysis unit and a data reliability analysis unit; the conventional data matching analysis unit performs similarity matching between conventional feature data in the current case and historical sample data in the database, and screens the historical sample data according to the matching results; according to the analysis results, sample data corresponding to the maximum matching degree between each type of conventional feature data and sample data is obtained in descending order, and the sample data is extracted and a comparison sample set is constructed; the relevant data extraction analysis unit performs relevant verification of each type of structured data by analyzing the difference feature data between each sample data in the comparison sample set and the current case. According to the calculation result of the degree of fit between each difference feature data and each sample data in the comparison sample set, the sample data in the comparison sample set corresponding to the maximum value of the calculated degree of fit of each difference feature data is selected, and the corresponding relevant evidence data, discretionary data and handling opinion data are collected for each sample data, and the collected data of each sample data are comprehensively acquired to obtain comprehensive relevant evidence group data, discretionary group data and handling opinion group data; the data reliability analysis unit performs reliability analysis on the current case based on the sample data corresponding to the acquired comprehensive relevant evidence group data, discretionary group data and handling opinion group data.
[0028] The report output module includes a report intelligent typesetting unit and a comprehensive output unit; the report intelligent typesetting unit intelligently typesets the relevant evidence, discretion and handling opinion data taken by matching the current case with the historical sample data and the paragraph text of the current case to generate an investigation termination report corresponding to the current case; the comprehensive output unit outputs the investigation termination report corresponding to the current case and its corresponding reliability analysis results and prompts the reviewer to confirm.
[0029] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the present invention combines multiple module units, and realizes the automatic generation of investigation termination reports by retrieving historical sample data from the database for corresponding model data training processing; it performs dual feature analysis by analyzing the difference features and conventional features of the target case and the historical sample report, and then extracts relevant data for each type of data by matching screening and matching analysis of the historical sample report, and finally realizes the quality analysis of the generated investigation termination report by analyzing the combined reliability of the extracted data and the target case, and outputs it to the reviewer through automatic typeset layout; the present invention greatly improves the efficiency of case handlers in writing investigation termination reports, and at the same time, combined with big data analysis, greatly reduces the information query steps of case handlers, and improves the reliability of the report, thereby improving the overall quality of report writing by case handlers. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] 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:
[0031] Figure 1 It is a structural schematic diagram of a system for improving the quality of investigation final report based on a digital model according to the present invention. DETAILED DESCRIPTION
[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0033] See also Figure 1 , the present invention provides a technical solution:
[0034] A method for improving the quality of investigation final report based on a digital model, the method comprising the following steps:
[0035] S100, automatically segment the historical investigation final report into paragraphs, send the text segmentation results to the model for training after manual review, extract report entity information from the investigation final report text, and integrate the extracted information to generate paragraph text;
[0036] S200, extracting features from the structured data of the current case according to the paragraph text data, and obtaining feature group data of the paragraph text of the investigation finalization report corresponding to the current case;
[0037] S300, performing association combination analysis based on the feature group data of the paragraph text of the current case in combination with the historical training data, obtaining the relevant evidence group data, discretion group data and handling opinion group data of the current case; and performing a comprehensive combination evaluation on the content text of the current case in combination with the obtained data;
[0038] S400: intelligently typeset the acquired data in combination with the paragraph text of the case according to the evaluation result and output the investigation completion report of the current case.
[0039] The S100 automatically segments the historical investigation final report into paragraphs, sends the text segmentation results to the model for training after manual review, extracts report entity information from the investigation final report text, and integrates the extracted information to generate paragraph text. The specific steps are as follows:
[0040] S101, constructing a historical data sample by retrieving the historical survey final report data in the database; segmenting the historical sample data by using the automatic segmentation process; manually reviewing the segmented sample data, and inputting the data into the training model for training after passing the review;
[0041] S102, extracting entity information data of the investigation finalization report in combination with the sample cutting data, and integrating the retrieved entity information to generate corresponding investigation report paragraph texts.
[0042] The specific steps of performing feature extraction on the structured data of the current case according to the paragraph text data and obtaining the feature group data of the paragraph text of the investigation finalization report corresponding to the current case in S200 are as follows:
[0043] S201, extracting structured data of the current case based on the paragraph text data of the historical sample data; and performing feature processing on the extracted structured data, and constructing a corresponding feature matrix for each structured data Where n is the number of structured data type labels, is the evaluation characteristic value of the corresponding feature point after the characterization processing of each type of structured data, and m is the number of corresponding feature points after the characterization processing of each type of structured data; by comparing the sample paragraph data, the feature difference analysis of the current case structured data is performed, and the analysis formula is: where dov n is the difference between the sample data corresponding to each type of structured data in the current case; C n is the sample data feature matrix corresponding to each type of structured data of the current case; and are the feature vectors generated by the feature matrix of the current case and the feature matrix of the sample data respectively; J(A n ) and J(C n ) are the variance values of the elements in the current case feature matrix and the sample data feature matrix respectively; m(A n ) and m(C n ) are the number of elements in the current case feature matrix and the sample data feature matrix respectively;
[0044] S202, manually set the judgment threshold dov s , perform feature screening on various types of structured data of the current case. If dov n ≥dov s , then the corresponding type of structured data is judged to be the difference feature data of the current case; if dov n <dov s, then the corresponding type of structured data is judged to be the conventional feature data of the current case; according to the judgment result, the feature matrix corresponding to the structured data of the difference feature data of the current case is combined, and the combination result is Among them A D is the difference feature matrix after combined processing; is the feature matrix corresponding to the structured data corresponding to the difference feature data; c is the type of difference feature data in the corresponding n types of structured data; c is the type data label of the difference feature data.
[0045] The S300 performs an associated combination analysis based on the feature group data of the paragraph text of the current case in combination with the historical training data to obtain the relevant evidence group data, discretion group data and handling opinion group data of the current case; the specific steps of performing a comprehensive combination evaluation on the content text of the current case in combination with the obtained data are as follows:
[0046] S301: Perform similarity matching between the conventional feature data in the current case and the historical sample data in the database, and filter the historical sample data according to the matching results. The matching analysis formula is: Among them, Sim i is the matching degree of the conventional feature data and sample data in the current case; Calculates vector dot product; The feature vector generated by the feature matrix corresponding to each conventional feature data in the sample data; is the feature vector generated by the feature matrix corresponding to each conventional feature data in the current case, where i is the number label of the conventional feature data type, i=nc; according to the analysis results, the sample data corresponding to the maximum matching degree between each type of conventional feature data and the sample data is obtained in descending order, and the sample data is extracted to construct a comparison sample set [B1, B2, B3...B i ]; among them B i The sample data corresponding to the maximum matching degree of each type of conventional feature data collected;
[0047] S302, by analyzing and comparing the difference feature data of each sample data in the sample set and the current case, the relevant evidence group data, discretion group data and processing opinion group data corresponding to each type of structured data are calculated, and the calculation formula is: in The degree of fit between each difference feature data and each sample data in the comparison sample set; The feature vector generated by the feature matrix corresponding to each difference feature data of the current case; is the eigenvector generated by the eigenmatrix corresponding to each difference feature data of each sample data in the comparison sample set; α and β are the horizontal angles of the eigenvector generated by the eigenmatrix corresponding to each difference feature data of each sample data in the current case and the comparison sample set; according to the calculation results of the fit between each difference feature data and each sample data in the comparison sample set, select the sample data in the comparison sample set corresponding to the maximum value of the calculated fit of each difference feature data, collect the corresponding relevant evidence data, discretion data and handling opinion data for each sample data, and coordinate the collected data of each sample data to obtain comprehensive relevant evidence group data, discretion group data and handling opinion group data;
[0048] S303: Perform reliability analysis on the current case based on the obtained comprehensive relevant evidence group data, discretion group data, and processing opinion group data corresponding to the sample data. The calculation formula is: Where e is the number of sample data in the comparative sample set that collects relevant evidence, discretion, and processing opinion data; The feature vector generated for the feature matrix of sample data in the comparison sample set corresponding to the collection of relevant evidence, discretion and processing opinion data.
[0049] The specific steps of S400 combining the acquired data with the paragraph text of the case for intelligent typesetting and outputting the investigation finalization report of the current case according to the evaluation result are as follows:
[0050] S401, intelligently typeset the relevant evidence, discretion and handling opinion data obtained by matching the current case with the historical sample data and the paragraph text of the current case to generate an investigation completion report corresponding to the current case;
[0051] S402: Output the investigation completion report corresponding to the current case and its corresponding reliability analysis results and prompt the reviewer to confirm.
[0052] A system for improving the quality of investigation final reports based on a digital model, the system comprising a model processing module, a feature data processing module, a report comprehensive analysis module and a report output module;
[0053] The model processing module automatically segments the historical investigation termination report into paragraphs, and after manual review, the text segmentation results are sent to the model for training, the report entity information is extracted from the investigation termination report text, and the extracted information is integrated to generate paragraph text; the feature data processing module performs feature extraction on the current case structured data based on the paragraph text data, and obtains the feature group data of the investigation termination report paragraph text corresponding to the current case; the report comprehensive analysis module performs association combination analysis based on the feature group data of the current case paragraph text in combination with the historical training data, and obtains the relevant evidence group data, discretion group data and handling opinion group data of the current case; the content text of the current case is comprehensively evaluated in combination with the acquired data; the report output module combines the acquired data with the case paragraph text for intelligent typeset according to the evaluation results and outputs the investigation termination report of the current case.
[0054] The model processing module includes a historical data cutting unit and an entity data extraction unit; the historical data cutting unit constructs a historical data sample by retrieving the historical investigation termination report data from the database; the historical sample data is segmented by using automatic paragraph segmentation processing; the segmented sample cutting data is manually reviewed, and after the review, the data is input into the training model for training; the entity data extraction unit extracts the entity information data of the investigation termination report in combination with the sample cutting data, and integrates the retrieved entity information to generate corresponding investigation report paragraph texts.
[0055] The feature data processing module includes a data difference analysis unit and a difference data judgment unit; the data difference analysis unit extracts structured data of the current case based on the paragraph text data of the historical sample data; and performs feature processing on the extracted structured data, and constructs a corresponding feature matrix for each structured data; performs feature difference analysis on the structured data of the current case by comparing the sample paragraph data; the difference data judgment unit performs feature screening on various types of structured data of the current case by manually setting a judgment threshold; and according to the judgment result, combines and processes the feature matrices corresponding to the structured data that matches the difference feature data of the current case.
[0056] The report comprehensive analysis module includes a conventional data matching analysis unit, a relevant data extraction analysis unit and a data reliability analysis unit; the conventional data matching analysis unit performs similarity matching between conventional feature data in the current case and historical sample data in the database, and screens the historical sample data according to the matching results; according to the analysis results, sample data corresponding to the maximum matching degree between each type of conventional feature data and sample data is obtained in descending order, and the sample data is extracted and a comparison sample set is constructed; the relevant data extraction analysis unit performs relevant verification of each type of structured data by analyzing the difference feature data between each sample data in the comparison sample set and the current case. According to the calculation result of the degree of fit between each difference feature data and each sample data in the comparison sample set, the sample data in the comparison sample set corresponding to the maximum value of the calculated degree of fit of each difference feature data is selected, and the corresponding relevant evidence data, discretionary data and handling opinion data are collected for each sample data, and the collected data of each sample data are comprehensively acquired to obtain comprehensive relevant evidence group data, discretionary group data and handling opinion group data; the data reliability analysis unit performs reliability analysis on the current case based on the sample data corresponding to the acquired comprehensive relevant evidence group data, discretionary group data and handling opinion group data.
[0057] The report output module includes a report intelligent typesetting unit and a comprehensive output unit; the report intelligent typesetting unit intelligently typesets the relevant evidence, discretion and handling opinion data taken by matching the current case with the historical sample data and the paragraph text of the current case to generate an investigation termination report corresponding to the current case; the comprehensive output unit outputs the investigation termination report corresponding to the current case and its corresponding reliability analysis results and prompts the auditor to confirm;
[0058] In the example:
[0059] Currently, a case handler needs to write an investigation termination report for a certain case, and the writing system is configured with the investigation termination report quality improvement system based on a digital model of the present invention; the historical data sample is constructed by retrieving the historical investigation termination report data in the database; the historical sample data is segmented by using the automatic segmentation process; the segmented sample segmentation data is manually reviewed, and after the review, the data is input into the training model for training; the sample segmentation data is combined with the sample segmentation data to extract the entity information data of the investigation termination report, and the corresponding investigation report paragraph texts are generated respectively according to the retrieved entity information.
[0060] According to the paragraph text data of the historical sample data, the structured data of the current case is extracted; the extracted structured data is characterized, and the corresponding feature matrix is constructed for each structured data. By comparing the sample paragraph data, the feature difference analysis of the current case structured data is performed. The analysis formula is: By manually setting the judgment threshold dov s , perform feature screening on various types of structured data of the current case. If dov n ≥dov s , then the corresponding type of structured data is judged to be the difference feature data of the current case; if dov n <dov s , then the corresponding type of structured data is judged to be the conventional feature data of the current case; according to the judgment result, the feature matrix corresponding to the structured data of the difference feature data of the current case is combined, and the combination result is
[0061] Based on the similarity matching between the conventional feature data in the current case and the historical sample data in the database, the historical sample data is screened according to the matching results. The matching analysis formula is: According to the analysis results, the sample data corresponding to the maximum matching degree between each type of conventional feature data and the sample data is obtained in descending order, and the sample data is extracted and a comparison sample set [B1, B2, B3...B i ]; By analyzing and comparing the difference feature data of each sample data in the sample set and the current case, the corresponding evidence group data, discretion group data and processing opinion group data of each type of structured data are calculated. The calculation formula is: According to the calculation results of the degree of fit between each difference feature data and each sample data in the comparison sample set, the sample data in the comparison sample set corresponding to the maximum value of the degree of fit calculated for each difference feature data is selected, and the corresponding relevant evidence data, discretionary data and handling opinion data are collected for each sample data, and the collected data of each sample data are comprehensively obtained to obtain comprehensive relevant evidence group data, discretionary group data and handling opinion group data; reliability analysis is performed on the sample data corresponding to the obtained comprehensive relevant evidence group data, discretionary group data and handling opinion group data and the current case, and the calculation formula is:
[0062] The relevant evidence, discretion and handling opinion data taken from the match between the current case and historical sample data are intelligently typeset with the paragraph text of the current case to generate the investigation completion report corresponding to the current case; the investigation completion report corresponding to the current case and its corresponding reliability analysis results are output and the reviewer is prompted to confirm.
[0063] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0064] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for improving the quality of investigation final report based on a digital model, characterized by: The method comprises the following steps: S100, automatically segment the historical investigation final report into paragraphs, send the text segmentation results to the model for training after manual review, extract report entity information from the investigation final report text, and integrate the extracted information to generate paragraph text; S200, extracting features from the structured data of the current case according to the paragraph text data, and obtaining feature group data of the paragraph text of the investigation finalization report corresponding to the current case; The specific steps of performing feature extraction on the structured data of the current case according to the paragraph text data and obtaining the feature group data of the paragraph text of the investigation finalization report corresponding to the current case in S200 are as follows: S201, extracting structured data of the current case based on the paragraph text data of the historical sample data; and performing feature processing on the extracted structured data, and constructing a corresponding feature matrix for each structured data Where n is the number of structured data type labels, is the evaluation characteristic value of the corresponding feature point after the characterization processing of each type of structured data, and m is the number of corresponding feature points after the characterization processing of each type of structured data; by comparing the sample paragraph data, the feature difference analysis of the current case structured data is performed, and the analysis formula is: where dov n is the difference between the sample data corresponding to each type of structured data in the current case; C n is the sample data feature matrix corresponding to each type of structured data of the current case; and are the feature vectors generated by the feature matrix of the current case and the feature matrix of the sample data respectively; J(A n ) and J(C n ) are the variance values of the elements in the current case feature matrix and the sample data feature matrix respectively; m(A n ) and m(C n ) are the number of elements in the current case feature matrix and the sample data feature matrix respectively; S202, manually set the judgment threshold dov s , perform feature screening on various types of structured data of the current case. If dov n ≥dov s , then the corresponding type of structured data is judged to be the difference feature data of the current case; if dov n <dov s , then the corresponding type of structured data is judged to be the conventional feature data of the current case; according to the judgment result, the feature matrix corresponding to the structured data of the difference feature data of the current case is combined, and the combination result is Among them A D is the difference feature matrix after combined processing; is the feature matrix corresponding to the structured data corresponding to the difference feature data; c is the type of difference feature data in the corresponding n types of structured data; c is the type data label of the difference feature data; S300, performing association combination analysis based on the feature group data of the paragraph text of the current case combined with historical training data, obtaining relevant evidence group data, discretion group data and handling opinion group data of the current case; and performing a comprehensive combination evaluation of the content text of the current case combined with the obtained data; The S300 performs an associated combination analysis based on the feature group data of the paragraph text of the current case in combination with the historical training data to obtain the relevant evidence group data, discretion group data and handling opinion group data of the current case; the specific steps of performing a comprehensive combination evaluation on the content text of the current case in combination with the obtained data are as follows: S301: Perform similarity matching between the conventional feature data in the current case and the historical sample data in the database, and filter the historical sample data according to the matching results. The matching analysis formula is: Among them, Sim i is the matching degree of the conventional feature data and sample data in the current case; Calculates vector dot product; The feature vector generated by the feature matrix corresponding to each conventional feature data in the sample data; is the feature vector generated by the feature matrix corresponding to each conventional feature data in the current case, where i is the number label of the conventional feature data type, i=nc; according to the analysis results, the sample data corresponding to the maximum matching degree between each type of conventional feature data and the sample data is obtained in descending order, and the sample data is extracted to construct a comparison sample set [B1, B2, B3...B i ]; among them B i The sample data corresponding to the maximum matching degree of each type of conventional feature data collected; S302, by analyzing and comparing the difference feature data of each sample data in the sample set and the current case, the relevant evidence group data, discretion group data and processing opinion group data corresponding to each type of structured data are calculated, and the calculation formula is: in The degree of fit between each difference feature data and each sample data in the comparison sample set; The feature vector generated by the feature matrix corresponding to each difference feature data of the current case; is the eigenvector generated by the eigenmatrix corresponding to each difference feature data of each sample data in the comparison sample set; α and β are the horizontal angles of the eigenvector generated by the eigenmatrix corresponding to each difference feature data of each sample data in the current case and the comparison sample set; according to the calculation results of the fit between each difference feature data and each sample data in the comparison sample set, select the sample data in the comparison sample set corresponding to the maximum value of the calculated fit of each difference feature data, collect the corresponding relevant evidence data, discretion data and handling opinion data for each sample data, and coordinate the collected data of each sample data to obtain comprehensive relevant evidence group data, discretion group data and handling opinion group data; S303: Perform reliability analysis on the current case based on the obtained comprehensive relevant evidence group data, discretion group data, and processing opinion group data corresponding to the sample data. The calculation formula is: Where e is the number of sample data in the comparative sample set that collects relevant evidence, discretion, and processing opinion data; The feature vectors generated for the feature matrix of the sample data in the comparative sample set corresponding to the collection of relevant evidence, discretion and processing opinion data; S400: intelligently typeset the acquired data in combination with the paragraph text of the case according to the evaluation result and output the investigation completion report of the current case.
2. According to the method for improving the quality of the investigation final report based on the digital model according to claim 1, it is characterized by: The S100 automatically segments the historical investigation final report into paragraphs, sends the text segmentation results to the model for training after manual review, extracts report entity information from the investigation final report text, and integrates the extracted information to generate paragraph text. The specific steps are as follows: S101, constructing a historical data sample by retrieving the historical survey final report data in the database; segmenting the historical sample data by using the automatic segmentation process; manually reviewing the segmented sample data, and inputting the data into the training model for training after passing the review; S102, extracting entity information data of the investigation finalization report in combination with the sample cutting data, and integrating the retrieved entity information to generate corresponding investigation report paragraph texts.
3. According to the method for improving the quality of the investigation final report based on the digital model according to claim 2, it is characterized by: The specific steps of S400 combining the acquired data with the paragraph text of the case for intelligent typesetting and outputting the investigation finalization report of the current case according to the evaluation result are as follows: S401, intelligently typeset the relevant evidence, discretion and handling opinion data obtained by matching the current case with the historical sample data and the paragraph text of the current case to generate an investigation completion report corresponding to the current case; S402: Output the investigation completion report corresponding to the current case and its corresponding reliability analysis results and prompt the reviewer to confirm.
4. A system for improving the quality of investigation final report based on a digital model, applying a method for improving the quality of investigation final report based on a digital model as claimed in any one of claims 1 to 3, characterized in that: The system includes a model processing module, a feature data processing module, a report comprehensive analysis module and a report output module; The model processing module automatically segments the historical investigation final report into paragraphs, and after manual review, the text segmentation results are sent to the model for training, the report entity information is extracted from the investigation final report text, and the extracted information is integrated to generate paragraph text; The feature data processing module extracts features from the structured data of the current case according to the paragraph text data, and obtains feature group data of the paragraph text of the investigation finalization report corresponding to the current case; The report comprehensive analysis module performs an associated combination analysis based on the feature group data of the paragraph text of the current case in combination with the historical training data to obtain the relevant evidence group data, discretion group data and handling opinion group data of the current case; combines the acquired data to perform a comprehensive combination evaluation on the content text of the current case; the report output module combines the acquired data with the paragraph text of the case for intelligent typeset according to the evaluation results and outputs the investigation termination report of the current case.
5. According to claim 4, a system for improving the quality of investigation final report based on a digital model is characterized in that: The model processing module includes a historical data cutting unit and an entity data extraction unit; the historical data cutting unit constructs a historical data sample by retrieving the historical investigation termination report data from the database; the historical sample data is segmented by using automatic paragraph segmentation processing; the segmented sample cutting data is manually reviewed, and after the review, the data is input into the training model for training; the entity data extraction unit extracts the entity information data of the investigation termination report in combination with the sample cutting data, and integrates the retrieved entity information to generate corresponding investigation report paragraph texts.
6. According to claim 5, a system for improving the quality of investigation final report based on a digital model is characterized by: The feature data processing module includes a data difference analysis unit and a difference data judgment unit; the data difference analysis unit extracts structured data of the current case based on the paragraph text data of the historical sample data; and performs feature processing on the extracted structured data, and constructs a corresponding feature matrix for each structured data; performs feature difference analysis on the structured data of the current case by comparing the sample paragraph data; the difference data judgment unit performs feature screening on various types of structured data of the current case by manually setting a judgment threshold; and according to the judgment result, combines and processes the feature matrices corresponding to the structured data that matches the difference feature data of the current case.
7. The system for improving the quality of investigation final report based on a digital model according to claim 6 is characterized by: The report comprehensive analysis module includes a conventional data matching analysis unit, a relevant data extraction analysis unit and a data reliability analysis unit; the conventional data matching analysis unit performs similarity matching between conventional feature data in the current case and historical sample data in the database, and screens the historical sample data according to the matching results; according to the analysis results, sample data corresponding to the maximum matching degree between each type of conventional feature data and sample data is obtained in descending order, and the sample data is extracted and a comparison sample set is constructed; the relevant data extraction analysis unit performs relevant verification of each type of structured data by analyzing the difference feature data between each sample data in the comparison sample set and the current case. According to the calculation result of the degree of fit between each difference feature data and each sample data in the comparison sample set, the sample data in the comparison sample set corresponding to the maximum value of the calculated degree of fit of each difference feature data is selected, and the corresponding relevant evidence data, discretionary data and handling opinion data are collected for each sample data, and the collected data of each sample data are comprehensively acquired to obtain comprehensive relevant evidence group data, discretionary group data and handling opinion group data; the data reliability analysis unit performs reliability analysis on the current case based on the sample data corresponding to the acquired comprehensive relevant evidence group data, discretionary group data and handling opinion group data.
8. According to claim 7, a system for improving the quality of investigation final report based on a digital model is characterized by: The report output module includes a report intelligent typesetting unit and a comprehensive output unit; the report intelligent typesetting unit intelligently typesets the relevant evidence, discretion and handling opinion data taken by matching the current case with the historical sample data and the paragraph text of the current case to generate an investigation termination report corresponding to the current case; The comprehensive output unit outputs the investigation finalization report corresponding to the current case and its corresponding reliability analysis result and prompts the auditor to confirm.
Citation Information
Patent Citations
Method, device and system for generating dynamic report
CN103020027A
File editing method and device
CN106021201A
Government affair industry knowledge base construction method based on natural language and user behavior analysis
CN115510025A
Method and device for converting document format
US20130174024A1
Method and system for determining structural blocks of a document
US20200034611A1
