Tibetan Entity Relationship Extraction Method

Abstract

The invention relates to a method for extracting a Tibetan entity relationship, which comprises the following steps: extracting training corpus from Tibetan-Chinese text corpus information; constructing a Tibetan word vector model; obtaining entity relationship feature vectors through the Tibetan word vector model; The relationship feature vector is used as input to build a neural network-based entity relationship classification model, and multi-layer feature extraction is performed on the entity relationship feature vector, and finally the Tibetan entity relationship classification is obtained. The present invention studies and solves the lexical semantic features and sentence feature vector representation methods of Tibetan entity relations by establishing a Tibetan word vector model, and then realizes the extraction of Tibetan entity relations by establishing a Tibetan entity relationship classification model, improving the Tibetan language The accuracy of entity relationship classification provides technical support and services for research in areas such as Tibetan knowledge graphs, question answering systems, information extraction, and information retrieval.

Description

technical field [0001] The invention relates to a method for extracting a Tibetan entity relationship, in particular to a method for extracting a Tibetan entity relationship based on a word vector. Background technique [0002] With the rapid popularization of the Internet, especially the rapid increase of Internet users in developing countries, the number of non-English text resources on the Internet has increased rapidly, and its growth rate has far exceeded the speed of 10 years ago. Published in multiple languages. According to a survey by the Minority Language Sub-Center of the National Language Resources Monitoring and Research Center of Minzu University of China: by the end of December 2011, the total number of websites in the mainland’s minority languages ​​was about 1,250, including 840 websites in Uyghur and 146 in Tibetan. and 136 Mongolian websites. "Compared with the growth rate of Internet users nationwide, the growth rate of Internet users of ethnic minoriti...

AI Technical Summary

Problems solved by technology

In Tibetan, usually called (Dalai Lama) for (The Dalai Lama), while current search engines do not show the relationship between the two

Moreover, all search results are mainly displayed in text containing keywords, without knowledge structure representation

Therefore, we cannot get comprehensive and accurate relevant information, let alone realize the in-depth mining of information

Images