CN-120804810-B - Processing method and system for improving recognition processing efficiency based on automatic key point grouping

CN120804810BCN 120804810 BCN120804810 BCN 120804810BCN-120804810-B

Abstract

The application relates to a processing method and a processing system for improving recognition processing efficiency based on automatic key point grouping, which belong to the technical field of content recognition processing, wherein the method comprises the steps of receiving a processing instruction proposed by a user, analyzing all files contained in the processing instruction to obtain a checking key point list and a corresponding grouping thinking chain; inputting all the examination points in the examination point list and the corresponding grouping thinking chain into a preset grouping intelligent agent, classifying all the examination points into a plurality of first-stage examination groups through the grouping intelligent agent, respectively distributing examination models for each first-stage examination group through an asynchronous call framework, generating and starting a plurality of examination threads, wherein the examination threads are in one-to-one correspondence with the first-stage examination groups for processing all examination points contained in the corresponding first-stage examination groups, and acquiring and merging and outputting examination results of all the examination threads for a user to acquire output results. The application has the effect of improving the processing efficiency of multiple files.

Inventors

ZHU XIAO
XU YANG
ZHOU LIHONG
ZHANG HUI
LU YIKAI

Assignees

苏州中格软件有限公司

Dates

Publication Date: 20260508
Application Date: 20250624

Claims (8)

1. A processing method for improving recognition processing efficiency based on automatic key point grouping is characterized by comprising the following steps: Receiving a processing instruction proposed by a user, and analyzing all files contained in the processing instruction to obtain an examination gist list and a corresponding grouping thinking chain; inputting all the examination points in the examination point list and the corresponding grouping thinking chains into a preset grouping agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping agent; an inspection model is distributed for each first-level inspection group through an asynchronous call frame, a plurality of inspection threads are generated and started, and the inspection threads are in one-to-one correspondence with the first-level inspection groups and are used for processing all inspection points contained in the corresponding first-level inspection groups; Obtaining and merging the examination results of all examination threads to obtain and output the output results; the examination key point at least comprises an entity to be examined; inputting all the examination points in the examination point list and the corresponding grouping thinking chains into a preset grouping agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping agent, wherein the method comprises the following steps of: Performing entity alignment processing on all the examination points in the examination point list, inputting the examination points subjected to the entity alignment processing and a corresponding grouping thinking chain into a preset grouping intelligent agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping intelligent agent, so that different entities correspond to different first-level examination groups, and generating a first mapping relation of the entities and the first-level examination groups; generating a secondary examination group by classifying examination points in each primary examination group based on a preset examination type, wherein the examination type at least comprises a plain text examination type and a visual image examination type; And respectively distributing a review model for each primary review group through an asynchronous call frame, generating and starting a plurality of review threads, wherein the review threads are in one-to-one correspondence with the primary review groups, and the method comprises the following steps of: matching the inspection models for each secondary inspection group respectively, generating and starting a plurality of inspection threads, wherein the inspection threads are in one-to-one correspondence with the secondary inspection groups; inputting all the examination points in the examination point list and the corresponding grouping thinking chains into a preset grouping agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping agent, and further comprising: analyzing the examination points contained in each secondary examination group, and determining a file list corresponding to each secondary examination group, wherein the file list contains files required to be called for executing the examination points contained in the corresponding secondary examination group; According to the second mapping relation, all examination points corresponding to the same file in each secondary examination group are respectively used as associated examination points, and three-level examination groups are formed by merging; Establishing association relation between three-level examination groups which belong to different two-level examination groups and correspond to the same file; the matching of the inspection models for each secondary inspection group respectively generates and starts a plurality of inspection threads, and the inspection threads are in one-to-one correspondence with the secondary inspection groups, comprising: Generating an inspection thread for each secondary inspection group, splitting the inspection thread corresponding to each secondary inspection group into sub-threads corresponding to each tertiary inspection group one by one according to the tertiary inspection groups contained in each secondary inspection group, and establishing an association relation between the corresponding sub-threads according to the association relation between the tertiary inspection groups; Respectively distributing an inspection model and an inspection engine for each sub-thread through an asynchronous call frame, and controlling the corresponding inspection model to execute the corresponding sub-thread through the inspection engine according to a preset inspection sequence until the corresponding inspection thread is completed; and triggering and executing the sub-threads with association relation with the target sub-thread when executing the target sub-thread, wherein the target sub-thread is any sub-thread.
2. The method of claim 1, wherein the matching of the inspection models for each of the secondary inspection groups generates and starts a plurality of inspection threads, and the inspection threads are in one-to-one correspondence with the secondary inspection groups, further comprising: Classifying the sub-threads with the association relationship to form a sub-thread list, analyzing examination points contained in each sub-thread in the same sub-thread list, respectively determining the dependency relationship among all the sub-threads in each sub-thread list based on analysis results, determining the execution mode of all the sub-threads in each sub-thread list according to the dependency relationship, and determining the execution sequence for the sub-threads with the execution modes being sequential execution, wherein the execution modes at least comprise parallel execution or sequential execution; Generating a scheduling scheme for each sub-thread list according to the execution mode and the execution sequence; and when executing the target sub-thread, triggering and executing the sub-thread with the association relation with the target sub-thread at the same time, wherein the method comprises the following steps: When executing a target sub-thread, triggering and executing the sub-thread with an association relation with the target sub-thread according to a scheduling scheme corresponding to the target sub-thread.
3. The processing method for improving recognition processing efficiency based on the gist automatic grouping according to claim 2, characterized in that the method further comprises: analyzing the specific examination contents of examination points contained in all the sub-threads in the same sub-thread list, and judging the similarity of the specific examination contents of the examination points contained in different sub-thread lists; Establishing multiplexing relations among examination points with similarity meeting preset requirements in different sub-thread lists, determining the multiplexing examination points as multiplexing examination points, and merging the multiplexing examination points with the multiplexing relations to form a multiplexing examination point set; Analyzing distinguishing examination contents and similar contents among the multiplexing examination points which belong to different sub-thread lists and have multiplexing relations in each multiplexing examination point set, wherein the distinguishing examination contents at least comprise entities to be examined, and the similar contents are the same examination contents in the specific examination contents corresponding to the multiplexing examination points having multiplexing relations; generating a multiplexing template for each multiplexing examination key point set, wherein the multiplexing template comprises similar contents and different examination contents, and constructing a multiplexing model for executing the multiplexing template; When executing the target sub-thread, triggering and executing the sub-thread with the association relation with the target sub-thread, and further comprising: And if the target sub-thread comprises multiplexing examination points, calling a corresponding multiplexing model, executing the multiplexing examination points through the multiplexing model according to the difference examination contents corresponding to the target sub-thread, and executing all examination points of non-multiplexing examination points contained in the target sub-thread through the examination model corresponding to the target sub-thread.
4. The processing method for improving recognition processing efficiency based on the gist automatic grouping according to claim 1, characterized in that the method further comprises: Recording an inspection path of the multiplexing model for multiplexing inspection points, and outputting a natural language interpretation report, wherein the inspection path at least comprises multiplexing template content, corresponding distinguishing inspection content and position information of the distinguishing inspection content in a corresponding file.
5. The processing method for improving recognition processing efficiency based on the gist automatic grouping according to claim 1, characterized in that the method further comprises: receiving a correction instruction proposed by a user, and verifying the identity of the user, wherein the correction instruction at least comprises a multiplexing template to be corrected and correction contents for correcting corresponding distinguishing examination contents; and when the user identity verification passes, correcting the multiplexing template contained in the correction instruction based on the correction instruction.
6. A processing system for improving recognition processing efficiency based on automatic point grouping, applied to the processing method for improving recognition processing efficiency based on automatic point grouping according to claim 1, characterized in that the system comprises: The file identification analysis module (201) is used for receiving a processing instruction proposed by a user, and analyzing all files contained in the processing instruction to obtain an examination key point list and a corresponding grouping thinking chain; the examination key point classification module (202) is used for inputting all examination key points in the examination key point list and corresponding grouping thinking chains into a preset grouping agent, and classifying all examination key points into a plurality of first-level examination groups through the grouping agent; The multithreading synchronous starting module (203) is used for respectively distributing a checking model for each first-level checking group through an asynchronous calling framework, generating and starting a plurality of checking threads, wherein the checking threads are in one-to-one correspondence with the first-level checking groups and are used for processing all checking points contained in the corresponding first-level checking groups; And the processing result output module (204) is used for acquiring and merging and outputting the examination results of all examination threads so as to enable a user to acquire the output results.
7. A processing device for improving recognition processing efficiency based on automatic point grouping, characterized by comprising a memory and a processor, wherein the memory has stored thereon a computer program that can be loaded by the processor and that performs the method according to any of claims 1 to 5.
8. A computer readable storage medium, characterized in that a computer program is stored which can be loaded by a processor and which performs the method according to any of claims 1 to 5.

Description

Processing method and system for improving recognition processing efficiency based on automatic key point grouping Technical Field The application relates to the technical field of content identification processing, in particular to a processing method and a processing system for improving identification processing efficiency based on automatic grouping of points. Background With the development of informatization, the number of documents that enterprises and individuals need to process and review increases exponentially, and in particular, in the fields of law, finance, scientific research and government, the need for multiple document inspection is increasingly prominent. The traditional file processing and review mode mainly depends on manual reading and comparison of file contents, and when the number of the file materials facing the file is large and/or the number of the review points of the file materials is large, the manual processing mode is low in efficiency and easy to influence the accuracy of review processing, so that improvement is needed. Disclosure of Invention In order to improve the examination processing efficiency and accuracy, the application provides a processing method and a processing system for improving the recognition processing efficiency based on the automatic grouping of points. In a first aspect, the present application provides a processing method for improving recognition processing efficiency based on automatic point grouping, which adopts the following technical scheme: Receiving a processing instruction proposed by a user, and analyzing all files contained in the processing instruction to obtain an examination gist list and a corresponding grouping thinking chain; inputting all the examination points in the examination point list and the corresponding grouping thinking chains into a preset grouping agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping agent; an inspection model is distributed for each first-level inspection group through an asynchronous call frame, a plurality of inspection threads are generated and started, and the inspection threads are in one-to-one correspondence with the first-level inspection groups and are used for processing all inspection points contained in the corresponding first-level inspection groups; and obtaining and merging the examination results of all examination threads to obtain and output the output results for the user. By adopting the technical scheme, all acquired files are analyzed at first, all examination points (namely examination points contained in an examination point list) required to be processed for all the files are determined, when a plurality of examination points are adopted, dynamic layering grouping of examination tasks can be realized by combining a grouping thinking chain, and finally, efficient parallel examination is realized by calling an adaptive model through asynchronous multithreading, so that manual work is replaced, and examination efficiency is improved. Optionally, the review gist includes at least an entity to be reviewed; inputting all the examination points in the examination point list and the corresponding grouping thinking chains into a preset grouping agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping agent, wherein the method comprises the following steps of: Performing entity alignment processing on all the examination points in the examination point list, inputting the examination points subjected to the entity alignment processing and a corresponding grouping thinking chain into a preset grouping intelligent agent, and classifying all the examination points into a plurality of first-level examination groups through the grouping intelligent agent, so that different entities correspond to different first-level examination groups, and generating a first mapping relation of the entities and the first-level examination groups; generating a secondary examination group by classifying examination points in each primary examination group based on a preset examination type, wherein the examination type at least comprises a plain text examination type and a visual image examination type; And respectively distributing a review model for each primary review group through an asynchronous call frame, generating and starting a plurality of review threads, wherein the review threads are in one-to-one correspondence with the primary review groups, and the method comprises the following steps of: And matching the inspection models for each secondary inspection group respectively, generating and starting a plurality of inspection threads, wherein the inspection threads are in one-to-one correspondence with the secondary inspection groups. By adopting the technical scheme, the primary grouping of the inspection points (namely forming the first-stage inspection group) is rea