Basic Study on Automatic Extraction of Haiku Seasonal Words

Abstract

As a first step towards building automatic text evaluation technology using artificial intelligence, especially haiku evaluation technology, a system that extracts seasonal words from haiku is implemented. The conventional extraction method used partial matching with the seasonal word database on the Internet. This method cannot identify some seasonal words that are combinations of nouns and verbs, so improvements are needed. The implemented system was created to solve the problem of seasonal word identification using partial matching, and its application to automatic generation of critical sentences is also being considered in the future. To improve accuracy, morphological analysis and similarity searches using historical kana usage was employed in this system. As a result of the evaluation, it was found that the system could identify seasonal words more reliably than amateur students. However, this system still has some drawbacks and requires many improvements.