Chinese word sense disambiguation method based on graph convolutional neural network

Abstract

The invention relates to a Chinese word sense disambiguation method based on a graph convolutional neural network (GCN). According to the invention, firstly, Chinese corpora are preprocessed; word segmentation, part-of-speech tagging and semantic tagging processing are performed on statements, containing ambiguous words, of the training and testing corpora; a word sense disambiguation feature graph is constructed by taking sentences where ambiguous words are located and word forms, part-of-speech and semantics contained in the sentences as disambiguation features and nodes, and weights are embedded into the nodes and edges by using Word2Vec and Doc2Vec tools and point mutual information (PMI) and TF-IDF methods; and the GCN model is trained by using the training corpus, and thus optimizing the model; word sense disambiguation is performed on the test corpus by using the optimized GCN model, so that probability distribution of ambiguous vocabularies under each semantic category can be obtained; and the semantic class corresponding to the maximum probability value is judged as the semantic class of the ambiguous vocabulary. The invention has a good word sense disambiguation effect, and the real meaning of the ambiguous vocabulary is more accurately judged.

Description

Technical field: [0001] The invention relates to a Chinese word sense disambiguation method based on a graph convolutional neural network, which can be well applied in the field of natural language processing. Background technique: [0002] Word sense disambiguation is a fundamental research topic in the field of natural language processing. In natural language, polysemy often exists, which often brings some troubles to applications in text classification, machine translation, and information retrieval. According to the specific language environment of the context, finding out the true semantics of ambiguous words and improving the accuracy of word representation will bring better results to the above application fields. [0003] At present, some common algorithms are often used to disambiguate and classify ambiguous words, such as Naive Bayesian, K-means, classification methods based on association rules and artificial neural networks. However, these traditional algorithm...

AI Technical Summary

Problems solved by technology

However, these traditional algorithms have some shortcomings. They cannot fully extract disambiguation features or are limited to local disambiguation feature extraction, and the classification effect of the classifier is not very good.

Images