Chinese event trigger word extraction method and device

Abstract

The invention provides a Chinese event trigger word extraction method and device. The method comprises the following steps: preprocessing an input text; performing initial vectorization; utilizing a heterogeneous attention network, catching the features of neighbor nodes of the same type as the current node emphatically, and catching the features of neighbor nodes of different types from the current node emphatically; and inputting the output of the type attention network into a conditional random field, and outputting a labeling sequence to realize the extraction of the trigger word. According to the invention, full word segmentation and dependency syntactic analysis are combined, multiple pieces of semantic information of the words are fused into the characters by fusing the semantic source information of the words, and the ambiguous word segmentation problem in a trigger word extraction task and the semantic ambiguity problem of Chinese words are solved. According to the invention, the heterogeneous graph attention network comprising the node attention network and the type attention network is utilized, the features of the neighbor nodes in the heterogeneous graph can be caught emphatically, and the extraction accuracy of the Chinese event trigger word is improved.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, and in particular relates to a method and device for extracting Chinese event trigger words. Background technique [0002] As a part of information extraction, event extraction has practical significance in public opinion analysis, automatic question answering, knowledge reasoning, etc. Event extraction refers to requiring people to use manual or automatic methods to identify target-related trigger words from semi-structured and unstructured data. As the core word of the event, the trigger word determines the type of the event. As a subtask of the event extraction, the extraction of the trigger word of the story event has practical significance for in-depth research. The existing Chinese event trigger word extraction faces two major problems: ambiguous word segmentation and word semantic ambiguity. The existing Chinese event trigger word extraction technologies are mainly di...

AI Technical Summary

Problems solved by technology

The existing Chinese event trigger word extraction technologies are mainly divided into three types: one is to use traditional machine learning methods, the problem is that it relies too much on NLP tools during feature extraction, and it can only capture the display features in the sentence; the other is to use CNN, RNN and other neural networks and their various improved methods have the problem that they are based on fixed word segmentation, which cannot solve the problems of ambiguous word segmentation and word semantic ambiguity; the third is to use graph convolutional networks, graph attention networks, etc. The problem with the neural network method is that it only uses words to construct an isomorphic graph structure or uses words and word segmentation results to construct a heterogeneous graph structure, and then uses methods such as graph convolutional networks or graph attention networks to complete the Chinese event trigger words. Extraction, which does not solve the problem of semantic ambiguity of words

[0003] In summary, the existing Chinese trigger word extraction technology affects the representation of characters to a certain extent due to incomplete feature capture and ambiguity, which in turn affects the extraction effect of Chinese event trigger words.

Images