利用标注者一致性数据估计多义词义项区分度

Inter-annotator Agreement as Indicator of Sense Distinction of Polysemy

多义词义项区分度指的是在实际语料中,人们为多义词选择合适义项的难易程度。本文探讨了一种测量多义词义项区分度的方法,试图量化表示义项区分度。我们认为,这个问题的难点在于多义词词典定义的不确定性:词典在定义多义词时缺乏强一致性的标准,以及不同人对词典定义的理解是不同的。这使得一方面基于词典的研究结果总是受到来自词典定义的挑战,另一方面研究者本身的主观性也多少削弱了研究结果的可靠程度。本文利用标注者一致性作为衡量多义词义项区分度的指标。结果显示,标注者一致性可以有效地表示多义词义项区分度,并且对进一步分析多义词义项的区别特征有指示作用。而且,多义词义项区分度大小与词典定义的清晰程度以及词频高低无关,而与区别特征与义项之间的映射关系有关,这也能够解释词典编纂者与使用者之间的差异。

In this article,we propose a method used to assess the sense distinction of polysemy.We also investigate how linguistic features shape the distinction.Polysemy is one of the essential research topics in the field of lexical semantics.Sense distinction of polysemy is how difficult it is when people are trying to determine a particular sense for a polysemous word in running texts.Given that there are very few works engaging this issue,we believe that the challenges are the lack of consistency among sense definitions in dictionaries and the unpredictable ways for humans to comprehend the definitions.We try to use inter-annotator agreement as indicator of sense distinction.The results validate the method.The agreement value can make suggestions on which way we should do further analyses.Last but not least,we findthe determining factor of the agreement value is how the linguistic features that distinguish senses are realized in running texts,which,we believe,justifies the‘gap’between dictionary editors and users.