Document information extraction method and system based on text classification and reading comprehension

Abstract

The invention belongs to the technical field of information content processing, and in particular relates to a document information extraction method and system based on text classification and reading comprehension. The method includes S1, converting the document into a plain text format; S2, preprocessing the document to obtain input data; S3, generating the corresponding word vector, word vector and context vector, and splicing to obtain the spliced ​​vector; S4, if splicing The latter vector is an answerable type, which is used as the input of the next step; S5, the position of the most matching long label data is obtained; S6, the final output is the long entity field to be extracted. The system includes a text information intelligent extraction module, a data preprocessing module, a feature extraction module, a text classification module, a reading comprehension module, a long entity label data generation module, and a data post-processing module. The invention has the characteristics of greatly shortening the training and prediction time, and improving the accuracy and speed of the document extraction model when extracting fields.

Description

technical field [0001] The invention belongs to the technical field of information content processing, and in particular relates to a document information extraction method and system based on text classification and reading comprehension. Background technique [0002] In today's highly informatized office, corporate office employees spend nearly 1 / 3 of their daily time dealing with text. For example, legal personnel have to review a large number of contracts and draft agreements; accounting personnel have to review a large number of reports. This kind of work has the characteristics of high repetition and large workload, and the manual processing efficiency is low, and it is easy to cause huge irreparable losses due to mistakes. In recent years, with the application and development of machine learning and deep learning in the field of natural language processing, the intelligent document review system has entered a stage of rapid development. [0003] The document intellig...

AI Technical Summary

Problems solved by technology

[0016] The purpose of the present invention is to overcome the problems of model training difficulties, time-consuming increase and low extraction accuracy in existing document information extraction methods in the prior art, and provides a method that can greatly shorten training and prediction time, and improve document extraction. Document Information Extraction Method and System Based on Text Classification and Reading Comprehension Based on Accuracy and Speed ​​of Model in Field Extraction

Images