自然语言文本共指消解性能评测算法研究

Research on Natural Language Text Coreference Resolution Evaluation Algorithms

在自然语言文本处理中,共指研究处于基础且关键的一环,它的有效解决将为众多的语言工程处理问题提供有力支持,因此对于共指消解性能的评测就显得尤为重要。针对共指消解任务的几种评测算法进行了深入研究,分析各种方法优缺点,指出CEAF方法是目前缺乏用于指代任务处理开放语料情况下,较适合汉语自然语言文本共指消解性能评价的一种方法。

Coreference resolution is a basal and important task in Nature Language Processing （NLP）, its availability can give strongly the support for many language engineering applications. Meanwhile, performance evaluation methods and algorithms are crucial in coreference resolution tasks. This paper discusses in detail several current evaluation methods and algorithms, and analyzes their merits and shortcomings, points out that CEAF is better than other algorithms in Chinese natural language text coreference resolution processing under the situation that there are not mass open and sharable annotated corpus focusing on Chinese coreference relation recognition.