The Role Collocation in Corpus Word Sense Annotation

Annotating words' senses in a corpus plays an important role in discriminating the senses of polysemants. When sense annotation based on dictionary there will be two obstacles: The first one is the ambiguous of dictionary sense distinctions, the second one is that senses in dictionary cannot match all usages of polysemants in corpus. Basing on the sentiment annotations of International Chinese Education Dynamic Corpus, this paper designs two kinds of collocation: lexical collocation and grammatical collocation and brings up that grammatical collocation can help finding the senses which aren't in dictionary, while lexical collocation can be used to solve annotation disagreements. The result of our experiment confirms the hypothesis and shows that those two collocations can be used to improve the accuracy of word sense annotation.