昔話の構成を表現する機能列の推定手法とその評価

抄録

In this paper, we propose a method to estimate sequences of functions to help to analyze folk tales. Functions are proposed by a Russian folklorist, V. Propp, to analyze folk tales. Estimating sequences of functions is a problem of assigning a function to each sentence in a tale, where a function represents an act of some characters in a sentence. We propose a method of supervised learning for estimating sequences of functions by using a Conditional Random Field (CRF). Observation sequences of a CRF are features of sentences that consist of nouns, verbs, and structures of sentences. We report some results to estimate sequences of functions from 50 folk tales.