Disambiguation of Japanese Idiomatic Expressions Using Bias in Occurrence of Verbs
Hirofumi Funato, Takehiro Teraoka · 2020
Some idiomatic expressions have both an original meaning and a literal meaning. It is important in understanding the context to determine which meaning of an expression is used in a sentence. An OpenMWE corpus has been used to determine the meaning of Japanese idiomatic expressions, but methods that use this corpus are not able to determine the meaning of expressions that are not included in the corpus. The only way to solve the problem with these methods is to increase the number of corpus examples by hand, which is time-consuming. In this paper, we propose a method for determining the meaning of idiomatic expressions by using the frequency of the occurrence of verbs and their related nouns in a sentence. The system consists of two phases: a learning phase and a disambiguation phase, in which a decision tree model is constructed and used to determine whether an expression in a sentence is idiomatic or literal. We also tested a method using literal usage representation, which has been used in previous studies. Evaluation experiments show that the combination of our method and a method of previous studies had the highest F value (0.795). An advantage of this method is that it is able to determine even idiomatic expressions that are not included in the corpus.