| 研究実績の概要 |
For this project emotion intensity lexicons were enhanced using word embeddings. Using this method several literary corpora were annotated for emotion arcs with a total of more than 10,000 annotated books. We were able to show that the mood of a novel can be computationally detected with high accuracy using only the first 500 words of the first chapter.
Furthermore, we examined specifically shame and guilt as cultural concepts in English and Japanese to show that shame is often portrayed as a public experience and guilt as an emotion that encompasses both private elements, akin to sadness, and public aspects, such as the motivation to openly acknowledge a transgression.
The results were disseminated over 7 peer-reviewed publications.
|
