组块分析研究综述

Research Review on Chunk Parsing

组块分析是一种新的句法分析思路和策略,侧重于解决底层歧义问题。文章归纳分析了组块分析的不同任务和研究方法,指出统计和规则方法相结合常常能取得更好的效果,但基本名词短语,最长名词短语等相对较大的语块的识别效果仍然不能令人满意。未来需要考虑如何将语言学特征融入统计机器学习,并且更好地将统计和规则方法结合起来。

Chunk parsing is a new syntactic analysis strategy, focusing on solving the shallow ambiguities.The article summarizes the different tasks and research methods of chunk parsing, and points out that the combination of statistics and rule methods can achieve better results, but the recognition effect of large chunks such as basic noun phrases and longest noun phrases still cannot satisfactory.The future needs to consider how to integrate linguistic features into statistical machine learning and integrate statistics and rule methods better.