Information extraction method based on deep semantic comprehension

Abstract

The invention provides an information extraction method based on deep semantic comprehension, which comprises the following steps of: constructing a body and a basic relationship in the field, and manually labeling parts of corpora; processing the manually annotated corpora, identifying an entity type corresponding to a specific relationship, and mining new words and synonyms in the field at the same time; merging synonyms recognized in the sentences, abstracting the original sentences and making syntactic analysis; clustering the abstracted sentences into sentence templates, and performing template learning; making sentence template evaluation; and performing new relationship extraction on manually unlabeled corpora by utilizing the sentence template, and evaluating and filtering a new relationship. According to the method provided by the invention, the syntactic analysis result can be better utilized, so that the automatically mined template has higher-level abstraction and generalization capabilities.

Description

technical field [0001] The invention relates to an information extraction method, in particular to an information extraction method based on deep semantic understanding. Background technique [0002] With the popularization of digital technology and Internet technology, text information has grown explosively. How to organize text information reasonably and find important information quickly and conveniently has become an urgent problem to be solved. [0003] One of the common ways to organize text information is structured charts, and information extraction is a common method for structuring structured, semi-structured and unstructured data. [0004] Among the existing information extraction methods, many attempts have been made on structured and semi-structured data, and better results can be achieved. But for unstructured data, the current extraction methods have the following problems: [0005] 1. The supervision-based method requires a large amount of manual labeling ...

AI Technical Summary

Problems solved by technology

[0005] 1. The supervision-based method requires a large amount of manual labeling data, and the model generated based on a certain field cannot be directly used in other fields, and still requires a lot of labeling work, and the versatility is not strong;

[0006] 2. The recall rate of the results obtained based on the semi-supervised method is too low, and a large amount of sample data is usually required for model optimization;

[0007] 3. The method based on remote supervision not only has the same shortcomings as the semi-supervised method, but also needs to prepare the existing structured information in advance for the areas of concern, and then optimize the model

[0008] It can be seen that the current information extraction methods based on syntactic analysis have a low accuracy rate for long sentences and cannot be directly applied to other tasks.

Images