Logical Semantics Theory and Relevance Theory for Models of Dialogue Understanding
Mitsuhiro OKADA*, Hide ISHIGURO*, Yuji NISHIYAMA*
and Michiko SUDO**
*Keio University / **Juntendo University
The purpose of this research is to give a logical analysis on the
meaning theory for the dialogical setting. We consider the two-layer
semantics theory for understanding dialogical utterances. The first
level is denotational semantics theory such as Montague-semantics and
the second is procedural semantics theory such as relevance theory.
We have developed a Montague-style denotational semantics theory for Japanese, where speaker's mood-expressions or intensional expressions are represented by specially introduced modal operators. Those modal operators are interpreted in terms of possible world (Kripke) semantics. Here, a mood expression is interpreted as a set of possible worlds. Such a denotational interpretation is considered as the basic meaning of the expression, and is given independently of the utterance situation or utterance context or speaker's specific intention.
A specific meaning of an utterance at a specific context or a specific intention is explained in the second level of meaning theory, procedural theory of meaning, where the meaning of a mood expression is understood as a procedure to choose or focus on some specific possible worlds out of the basic denotational set of possible worlds (of the first semantics level). While a unique basic denotation is assigned to each expression (in principle) in the first (denotational) semantics level, many different ways of focusing on the denotation (possible worlds), hence many different procedural meanings, may be assigned to each (mood) expression, depending on the speaker's intention or context, in the second (procedural) semantics level.
Those specific possible worlds focused often represent the common knowledge or context of the utterance between the speaker and hearer: Further conversation is understood in such a way that the content is consistent (or in some relation) with the focused possible worlds (i.e., context information), while further focusing process of possible worlds is carried out by the procedural meaning of some utterances. The second level of semantics theory, hence, gives a dynamic view of development of dialogical communication processes.
The most important information with which a speaker makes a hearer (and a hearer makes himself/herself) understand speaker's specific intention (or specific procedural meaning) is (1) the possible worlds (the context information) so far specified, and (2) the voice information at the speaker's utterance. As the speaker's voice information, the most important are (1) intonation, (2) accent, (3) duration length of vowels. Our goal of research is to show how much those voice informations can characterize speaker's specific intentions (i.e., can distinguishes various different procedural meanings); in other words, to show how much the voice information contributes to the mutual understanding of a conversation in the above-described communication process.