Entity Translation Method Based on Web Retrieval

Abstract

The invention provides a web retrieval based entity translation method. The method comprises the steps that entity description information and an entity to be translated in a knowledge base are utilized for performing web retrieval; the entity description information in the knowledge base is used for performing sequence labeling on a historical retrieval result to obtain at least one candidate entity translation; according to a current character / word TF-IDF value in the historical retrieval result and the occurrence probability of the translation entity and the current character / word, at leastone candidate enhancing word is obtained; statistics is performed on the related statistic amount between the candidate entity translation and the candidate enhancing word, and a retrieval state table is generated or updated; the retrieval state table is adopted as a reinforcement learning state set, the at least one candidate enhancing word and special 'ending' motions are adopted as a reinforcement learning motion set, the optimal retrieval enhancing word selective strategy is obtained through a reinforcement learning mechanism, and the candidate entity translation with the highest occurrence frequency is adopted as the final entity translation in the end. The entity translation accuracy and translation efficiency can be improved, and meanwhile the problems of unlisted words and entityname ambiguity can be avoided.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a method for entity translation based on web retrieval. Background technique [0002] In recent years, with the emergence and development of a large number of open knowledge bases, the imbalance of knowledge base construction has become prominent. This imbalance is mainly reflected in: the coverage of different knowledge bases is not the same; the knowledge of different languages There are huge gaps in orders of magnitude between libraries. For the construction of a knowledge base in a new domain or language, entity translation technology has the advantages of rapid construction and excellent structural compatibility. [0003] The core of the construction of the translation knowledge base is the translation of knowledge base entities. However, due to the rich connotation of the entity concept, entity translation based on rules or statistical methods often en...

AI Technical Summary

Problems solved by technology

[0005] Because the retrieval results of web retrieval usually cannot return enough accurate information, the existing entity translation based on web retrieval has the defects of low translation accuracy and low translation efficiency

Images