中文摘要 |
當今蓬勃發展中的深度學習,是第三時期AI的堅固的磐石。應用深度學習來處理人類自然語料時,光只處理言語來企圖追求正確性是有其界限的。而AI是透過分析現實中,人類的多模式溝通語料來進行自然言語之學習。這一點與把日語當作第二外語來學習的狀況是相同,因此掌握具體言語表現運用的實況之研究,是有其必要性。本論文目的有三。第一是考察現今被常使用AI深層學習上之語料類型。第二是應用多模式分析方法,將語言表達與非語言表達間的聯繫,視為一個單位來掌握跟切割。第三是將上述成果當基底,來進行考察AI與人文社會學科研究以及日語教育之交互關係。 Currently, deep learning, which is the basis of the third stage development of AI; artificial intelligence, has limitations on the precision that can be obtained by processing only languages even when processing human natural language information. Through data analysis of real human multimodal communication, AI is learning about human natural language. This point is the same for human's Japanese learning, and Japanese language and education research is required to capture the actual state of specific language expression operation. In this thesis, we consider the data type in deep learning etc. of AI that is currently used, and come up with negotiation of expression of language and non-language as one expression unit by multimodal analysis method. Moreover, this thesis considers the relationship between humanities-social studies as well as Japanese language education and development of AI. |