A method for extracting specific text information based on layer classifier and template matching

Abstract

The invention discloses a method for extracting specific information by using a layer classifier and a template matching. The method constructs a set of information extraction models according to a mass of text data by using a layer classifier model, an information extraction template, a semantic tree and other technologies. In the process of constructing the model, firstly, the layer classifier and the trigger-based method are used to determine the type of the text, then the information extraction template shared by all the text can be used, and the corresponding information extraction template can be selected according to the type of the text, and finally the specific information can be extracted from the text data through the template. Through the invention, the specific information canbe extracted quickly and accurately from the massive text.

Description

technical field [0001] The invention relates to a text information extraction method, in particular to a specific text information extraction method based on layer classifiers and template matching. Background technique [0002] In order to extract specific information from massive text data, the original text data can be extracted through text information extraction technology, and then the specific information points extracted from the text can be integrated in a unified form. A successful information extraction system can convert massive unstructured text data into structured information data, and can also store the converted data into the database. This is of great significance for text analysis, public opinion monitoring, Internet knowledge acquisition and other fields. [0003] The specific information extraction system needs to adopt different templates for different text types. First, the text needs to be classified. However, when the text types increase rapidly, ...

AI Technical Summary

Problems solved by technology

However, when the text types increase rapidly, the classification effect of ordinary classifiers will continue to decrease as the number of types increases, resulting in a large number of misclassifications of text types

Secondly, after the type is determined, a rule-based method is often used to extract different information points, but purely rule-based extraction will lead to a large loss of connections between text syntax, thereby reducing the information extraction effect

Images