Method for automatically identifying named entities of traditional Chinese medicine patent literatures

Abstract

A method for automatically identifying named entities of traditional Chinese medicine patent literatures relates to the field of natural language processing and can be used for information retrieval of the traditional Chinese medicine patent literatures and construction of a traditional Chinese medicine knowledge graph. The method is characterized in that a delimiting module and a classification module are established based on a word n-ary, the delimiting module and the classification module are trained firstly, and then the trained delimiting module and classification module are used for automatically identifying named entities of traditional Chinese medicine patent literatures. The method has the following advantages: (1) traditional Chinese medicine named entity identification is directly carried out based on a Chinese character sequence without word segmentation; and (2) the traditional sequence labeling technology is not used, so that the generation of illegal labeled sequences isavoided.

Description

Technical field [0001] The invention relates to the field of natural language processing, and in particular to a method for automatically identifying named entities in Chinese medicine patent documents. Background technique [0002] Existing Chinese named entity automatic recognition methods are generally based on the following two technologies. (1) Named entity recognition based on Chinese words. This method needs to segment the Chinese text in advance. (2) Based on the traditional sequence labeling technology, this method labels BIO tags on each word, and then converts the BIO tags into named entity tags. The named entity recognition methods based on these two technologies have some shortcomings, which are described in detail as follows. [0003] First, the word segmentation-based named entity recognition method has the following disadvantages: (1) The Chinese word segmentation tool itself has inaccuracy and errors. These word segmentation errors will greatly affect the result o...

AI Technical Summary

Problems solved by technology

[0005] Aiming at some of the above-mentioned problems in traditional named entity recognition: 1. Word segmentation needs to be performed first, and word segmentation errors will be extended to named entity recognition, so that the named entities in the wrong part of word segmentation cannot be recognized correctly; 2. Based on Traditional sequence annotation techniques can generate illegal tagged sequences

Images