This paper presents a method of constructing a database of Japanese compound verb examples, and evaluates the database. The objective of constructing this database is to analyze the relationship between Japanese compound and component verbs. Whether to include a compound verb in the database is determined semi-automatically by the number of examples that can be extracted from these Web corpora. The actual database that resulted from this method consists of 3371 compound verbs (median number of examples per verb = 1173). It covers 77.2% of the relevant entry words in the Iwanami Japanese language dictionary. A comparison with a general-purpose Web corpus shows that this method enabled to collect more than 1000 examples for 1829 compound verbs with a wide range of probability of occurrence. The average cosine similarity between the distributions of case-marked elements in the database examples and in those extracted from the Web corpus is 0.878 for compound verbs. Therefore, this result suggests that the bias of examples is controlled.
『Webデータに基づく複合動詞用例データベース』の構築と評価
papers1702.pdf
(1.23MB)
概要
