A semantic role annotation-based document body cutting and classifying system and a semantic role annotation-based document body cutting and classifying method

Abstract

The invention discloses a document body cutting classification system and method based on semantic role annotation. The system comprises a semantic role annotation engine, a knowledge body library, abody cutting recognition rule engine and a body cutting recognition rule library which are connected. The semantic role labeling engine comprises word dividers which are connected with one another; wherein the body cutting recognition rule engine comprises a body cutting recognition rule parser, a body cutting recognition rule matcher and a body cutting recognition rule inference device which areconnected with one another, and the body cutting recognition rule parser is used for parsing a body cutting recognition rule into a data structure which can be recognized by a computer program from atext; the body cutting recognition rule matcher is used for matching a result labeled by the semantic role labeling engine with a body cutting recognition rule; and the body cutting recognition rule inference device is used for executing inference according to the matching result of the body cutting recognition rule matcher to obtain the final body cutting classification. The method is relativelylow in data cost, relatively low in calculation cost and relatively high in posture recognition accuracy.

Description

technical field [0001] The invention relates to the field of document genre classification, in particular to a document genre classification system and method based on semantic role labeling. Background technique [0002] When classifying party and government documents such as party documents, special policies, laws and regulations, and leadership speeches, the traditional method is to collect a large amount of corpus for different document genres, use machine learning training models, and classify documents based on the machine learning training model , the number of documents summarized is large, and the data cost is high. In addition, the complete document content needs to be processed, and the calculation cost is high. The accuracy of this genre identification method is low, because party and government documents are not as short as news information, and many party and government documents have dozens or even hundreds of pages. It is difficult for machine learning techn...

AI Technical Summary

Problems solved by technology

The accuracy of this genre identification method is low, because party and government documents are not as short as news information, and many party and government documents have dozens or even hundreds of pages.

It is difficult for machine learning techniques to accurately train accurate classification models

Images