This paper addresses the problem of choosing of a proper logical formalism for the underspecified semantic representation (USR) of NL utterances, and investigates the consequences of the idea that utterance meanings should be viewed as context-changing operations. Allan Ramsey argues against the need of a special dynamic logic, or of developing other special logics for modelling the USR behaviour. Particularly, concrete, but also general arguments are provided (e.g. the increased expressive power of the USR language involves a worse situation for the computational tractability of the problem). The author supports as the adequate logical formalism for dealing with USRs a doxastic logic that is a minor extension of first-order logic, allowing belief sets of various kinds to be represented by propositions and using a constructive interpretation of the underlying logic. The proposed doxastic logic accounts for the dynamics and underspecifications of the meaning of NL utterances, and is particularly suited for the standard theorem proving techniques.
Reviewer: Neculai Curteanu (Iaşi)