1994 年 9 巻 6 号 p. 899-907
An utterance has implied meanings. Hence, a computer has to infer implied meanings to communicate with humans smoothly. This paper deals with some of implied meanings which occur when a speaker gives a hearer an insufficient answer for a question. We assumes that the implied meanings are derived from an utterance with the following factors : relation between the utterance and its context, the speaker and hearer's knowledge of the utterance, their loss and gain, their evaluation of the utterance, their hierarchy, their friendship, and the form of the utterance. We give formal descriptions of the factors and their rules to obtain the implied meanings. The knowledge of an utterance is expressed with a modal logic based on KD45_<(M)>. The hearer infers the implied meanings of the speaker's insufficient answer with the following rules. (l) Default Unknown Rule : The speaker does not know the lacked information when it is not clear that the speaker knows the information. (2) Affect Ignorance Rule : The speaker affects ignorance of the lacked information if it is clear that the speaker knows the information and it is loss to him. (3) Embarrassed Rule : The speaker is embarrassed with the lacked information if it is clear that the speaker knows the information and it is gain to him. (4) Hesitation Rule : The speaker hesitates to speak the lacked information if it is clear that the speaker knows the information and it is loss to the hearer. (5) Default Negation Rule : The speaker negates the lacked information if it is clear that the speaker knows the information and it is neither loss nor gain to both the speaker and the hearer. (6) Unpleasantness Rule : The speaker feels unhappy to talk about the lacked information if the hearer cannot infer the lacked information with the above-mentioned rules.