A Method and System for Entity Alignment in Cross-lingual Knowledge Space Based on Link Prediction

Abstract

The present invention relates to a method and system for realizing entity alignment in a cross-language knowledge space based on link prediction, which combines cross-language knowledge space triples, knowledge space representation learning, predicts new aligned entity pairs, and adds new training data through self-learning. The first step is to generate predicted entity pairs. For a small amount of training corpus, a simple method based on link prediction is designed to predict new entity pairs, and cross-language knowledge space fusion is performed. On this basis, data fusion in the two knowledge spaces is improved. The efficiency of judging whether the entity pairs in the cross-lingual knowledge space are the same entity, the design of the method model is relatively light, and it saves the manpower of labeling.

Description

technical field [0001] The invention relates to the fields of artificial intelligence and machine learning, in particular to a method for realizing entity alignment in cross-language knowledge space based on link prediction. Background technique [0002] With the development of modern Internet technology, more and more information is accumulated on the Internet. When users retrieve information and consult e-commerce companies, they are not only satisfied with the results of traditional simple keyword search and matching, but also pursue more intelligent and personalized search and Q&A services. At present, a large number of Internet companies at home and abroad, such as Google, Amazon, Baidu, and Tencent, have established their own knowledge space systems, using knowledge space technology to provide customers with more intelligent services. Knowledge space technology can combine relevant knowledge in reality, No matter in the field of information retrieval or intelligent qu...

AI Technical Summary

Problems solved by technology

However, these knowledge spaces have certain limitations. For example, the English knowledge space does not include comprehensive information for non-English regions. Therefore, by integrating information from these different language knowledge spaces, a larger and more comprehensive multilingual knowledge space system is formed. , can provide customers with more comprehensive information services

However, the traditional word translation model-based fusion technology is limited by the accuracy of word translation and the polysemy of the word itself. For example, the word Chaoyang can refer to Chaoyang District in Beijing or Chaoyang City in Liaoning Province.

The deep learning model learns the entity and relationship vectors of the knowledge space, and uses the labeled entity pairs to train the alignment model, which requires labeling a lot of information and consumes a lot of manpower

Images