基于依存结构的语块分析

Research of text chunking on the basis of dependency structure

语块分析在自然语言处理研究中占有重要的地位。传统方法将语块分析看作是序列标注任务,但由于受到马尔可夫假设的限制,无法对较长语块进行分析建模。提出了基于依存结构的语块分析方法,通过建立语块的中心词和其他词之间的依存关系,方便了较长语块的分析建模。研究表明,该方法提升了较长语块的分析性能。

Text chunking had played an important role in the research of Natural Language Processing. Traditionally, chunking was regarded as a sequence labeling task. But due to the limit of Markov assumption, long chunks could not be easily modeled. This article proposed a chunking method based on dependency structure so as to model longer chunks through dependencies between chunks＇ heads and other words. Experimental results showed that such a method could improve long chunks＇ performances.