Chinese-cross-language word embedding method fusing word cluster constraints

Abstract

The invention relates to a Chinese cross-language word embedding method fusing word cluster constraints. The method comprises the following steps: firstly, preprocessing a Chinese-View lingual corpus and training Chinese-View lingual word embedding; then constructing a Chinese-Vietnamese bilingual dictionary and a word cluster dictionary by using an online dictionary and an open source word library, and fusing alignment information of two granularities of words and word clusters into a training process of a mapping matrix; and finally, obtaining a Chinese and Vietnamese shared word embedding space through a cross-language mapping framework, so that Chinese and Vietnamese words with the same meaning are embedded in the space to be close to each other. According to the method, the word cluster alignment information in the Chinese-View bilingual dictionary is extracted by using different types of association relationships, so that the mapping matrix learns a multi-granularity mapping relationship, the generalization of the mapping matrix on non-tagged words is improved, and the problem that the bilingual space alignment effect is poor in a Chinese-View low-resource scene is solved. Experimental results show that the alignment accuracy of the model on induction tasks (at) 1 and (at) 5 in the Chinese-cross dictionary is improved by 2.2 percentage points compared with that of a VecMap model.

Description

technical field [0001] The invention relates to a Chinese-Spanish cross-language word embedding method fused with word cluster constraints, and belongs to the technical field of natural language processing. Background technique [0002] Cross-language word embedding maps words with the same meaning in different languages ​​to the same space for alignment. It is the basis for tasks such as cross-language text classification, cross-language sentiment analysis, machine translation, and cross-language entity linking, and has important application value. [0003] Chinese-Vietnamese cross-language word embedding is a bilingual word embedding task for low-resource languages. At present, low-resource cross-language word embedding methods mainly include unsupervised, semi-supervised and supervised three categories. Unsupervised methods exploit the similarity of monolingual embedding spaces in different languages, and can learn mapping matrices to achieve alignment without labeling da...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a Chinese-Vietnamese cross-language word embedding method that integrates word cluster constraints to solve the problem of lack of large-scale The bilingual dictionary leads to the problem that the learned mapping matrix has weak generalization on non-marked words outside the dictionary, and cannot accurately align the bilingual word embedding space

Images