面向自然语言生成的词汇选择研究

Research on Lexical Choice for Natural Language Generation

词汇选择是自然语言生成研究的关键问题,主要研究如何将意思表示映射为恰当词语。目前许多自然语言处理系统仍没有采取恰当的选词策略,选词效果非常不理想。本文从词汇选择研究的理论和方法两个方面,对当前词汇选择研究现状进行梳理,重点分析了当前研究存在的重点和难点,介绍了现有研究在词义表示、词汇选择建模等方面的方法;并介绍了今后词汇选择研究可能的发展趋势。

Lexical choice is a key issue in natural language generation research, which mainly studies how to map meaning representations to appropriate words. At present, many natural language processing systems still do not adopt an appropriate lexical choice strategy, and the lexical choice effect is very unsatisfactory. This paper sorts out the current research status of lexical choice from two aspects of the theory and method, focuses on analyzing the key points and difficulties, and introduces the existing research methods in word meaning representation, lexical modeling, etc.. Finally, we introduce the possible development trend of lexical choice research in the future.