7 0 0 0 OA 「しりとり」に含まれる手がかりが「準」自然言語の意味推測に与える影響
- 著者
- 勝 将也 中島 綾乃 菊池 華世 中島 亮一 大澤 正彦
- 出版者
- 日本認知科学会
- 雑誌
- 認知科学 (ISSN:13417924)
- 巻号頁・発行日
- vol.30, no.3, pp.314-326, 2023-09-01 (Released:2023-09-15)
- 参考文献数
- 36
It is important to maintain the communication with an agent (e.g., a robot), when people enjoy the human-agent communication. Recent studies reported that using shiritori, a game where players say a word starting with the last letter of the previous word, may be effective to verbally communicate with an agent speaking semi-natural language (i.e., words expressed by the combination of sounds of “do” and “ra”). This study examined what factors influence the communication with the agent using such language. Experiment 1 investigated the effect of the timing of providing information on the guessing of the word expressed by semi-natural language. Participants watched a video clip of the agent using such language, guessed the meaning of the word it spoke, and reported their confidence in their guess. They were provided information about the initial letter and the number of characters before or after watching the video clips. The results suggest that the timing of the information is not important to guess the semi-natural language word. Experiment 2 investigated the effect of shiritori with the agent on the guessing of semi-natural language. Participants were assigned to one of three groups: assuming shiritori, informed of the initial letter, and non-informed groups. The confidence rating was higher in the shiritori group than in the other groups. Therefore, the information provided by shiritori should be important to guess the semi-natural language words. We also discussed the possibility that the typical word pairs in shiritori can influence the guessing of such language.
- 著者
- 大澤 正彦
- 出版者
- 一般社団法人 人工知能学会
- 雑誌
- 人工知能 (ISSN:21882266)
- 巻号頁・発行日
- vol.34, no.1, pp.106-108, 2019-01-01 (Released:2020-09-29)
3 0 0 0 OA 全脳アーキテクチャ若手の会
1 0 0 0 OA 前頭前野のAccumulatorモデルを利用した,階層型脱抑制システムの提案
- 著者
- 芦原 佑太 大澤 正彦 島田 大樹 栗原 聡 今井 倫太
- 出版者
- 人工知能学会
- 雑誌
- 人工知能学会全国大会論文集 (ISSN:13479881)
- 巻号頁・発行日
- vol.31, 2017
前頭前野は,多くの脳領域の抑制・脱抑制や調停を行う機能を持つと考えられている.また,Accumulatorとして動作する神経細胞は前頭前野においても発見されており,特に自発的な運動の開始に関与しているという主張が報告されている. 本研究ではこれまで前頭前野に関して報告されている知見を参考に,Accumulatorモデルを用いて複数の機械学習器を調停する階層型脱抑制システムを提案する.