2021年度 人工知能学会全国大会(第35回)

講演情報

国際セッション

国際セッション(Work in progress) » EW-5 Human interface, education aid

[1N3-IS-5b] Human interface, education aid (2/2)

2021年6月8日(火) 15:20 〜 16:20 N会場 (IS会場)

Chair: Toshihiro Hiraoka (The University of Tokyo)

15:20 〜 15:40

[1N3-IS-5b-01] A Prototype System that Transforms Stories in the Level of Conceptual Elements Using a Noun Conceptual Dictionary

Based on the Theme of Love and Sex

〇Jumpei Ono1, Miku Kawai2, Takashi Ogata2 (1. Aomori University (Tokyo Campus), 2. Iwate Prefectural University)

キーワード:Narrative Generation, Love and Sex Story, Noun Concepts, Noun Conceptual Dictionary

The problems of love and sex with robots have been studied in the “International Congress on Love & Sex with Robots” organized by David Levy, who is a chess player and is an author of books of AI. Of course, also in narratives and stories, love and sex are important theme. The objective of this paper is to present a prototyping system, based on the themes of love and sex, that generates a new story representation through exchanges the concepts, including characters, objects, and locations, in an inputted story representation with new concepts using the noun conceptual dictionary developed by us.
Although new noun concepts to be exchanged are selected from the range of the concepts related to love and sex in the noun conceptual dictionary, the selection is done using three types of methods with: (1) random selection, (2) co-occurrence, and (3) hierarchical structure in the dictionary. For instance, the second method exchanges all of the corresponded concepts in the inputted story into new love and sex-related concepts in the noun conceptual dictionary using co-occurrence information. Furthermore, we conduct on a questionnaire survey for the generation results by the developed mechanisms. For example, the subjects compare an inputted story and the transformed stories, and evaluate the degree of the outputted stories are transformed to love and sex stories.

講演PDFパスワード認証
論文PDFの閲覧にはログインが必要です。参加登録者の方は「参加者用ログイン」画面からログインしてください。あるいは論文PDF閲覧用のパスワードを以下にご入力ください。

パスワード