人工知能
Online ISSN : 2435-8614
Print ISSN : 2188-2266
人工知能学会誌(1986~2013, Print ISSN:0912-8085)
Frustration-Based Learning法による隠喩の理解
諏訪 正樹元田 浩
著者情報
解説誌・一般情報誌 フリー

1990 年 5 巻 3 号 p. 291-299

詳細
抄録

Metaphors are pervasive in speech and thinking activities in our daily life. Understanding a metaphorical sentence means to discover some general metaphors which underlie the sentence. In this paper, we claim that Frustration-Based Learning method (FBL method) can be applied to understanding metaphors. The method has been already proposed as a means of acquiring strategies for producing appropriate auxiliary-lines in auxiliary-line problems in elementary geometry. FBL method has two main concepts ; identification of frustrated states and limited forward reasoning within a restricted world. Identification of frustrated states contributes to discovering where to learn in the whole problem. If a metaphorical sentence is underlied by several general metaphors, several frustrated states are to be identified during the processing of the sentence, and the following learning tasks are performed for each frustrated state ; to suppose the restricted world which has been concerned with the resolution of each frustrated state and then to enumerate all the pieces of information which hold within the world. The latter task is called limited forward reasoning within a restricted world, which directly contributes to discovering underlying information of the sentence, instead of interpreting it superficially. FBL method is based on the concept that learning how frustrated states are resolved in a problem solving process leads to understanding the problem, and its application to metaphors results in understanding a sentence as a composition of several general metaphors.

著者関連情報
© 1990 人工知能学会
前の記事 次の記事
feedback
Top