Mark Seligman
CO-OC: Semi-automatic Production
of Resources for Tracking Morphological
and Semantic Co-occurrences
in Spontaneous Dialogues
Abstract:This document presents a tentative set of Communicative Acts (CAs) which has been used to label
spontaneous dialogues in both English and Japanese. A Communicative Act is a communicative goal
or aim which (according to native judgments) can be expressed in language L by a distinctive set of
conventional cue patterns in specified discourse contexts. Communicative Acts are thus similar to
speech acts, and similar to the pragmatic categories often called IFTs (illocutionary force types) at
ATR. However, we restrict our attention to communicative goals which can be explicitly expressed
via conventional surface cue patterns, thus excluding goals which are expressed using one-time-only
combinations, goals which are expressed only implicitly, or goals which can only be defined in terms
of relations between utterances. We describe methods of discovering and revising CAs which depend
on native judgments concerning essential equivalence of meanings and functions of cue patterns in
context. While these judgments are subjective, they concern shared conventions regarding objectively
observable objects — the cue patterns and contexts. Thus a consensus can be expected to emerge
during repeated revision. In this important respect, the methodology is data-driven or corpus-based.
The present study also emphasizes comparison of CAs in English and Japanese. We find that most
of our proposed CAs are valid for both English and Japanese: only two out of 27 CAs seem to be
monolingual for our corpus. We begin by introducing CAs and briefly describing the background,
goals, and status of our research. In later sections, we discuss our methodology in greater depth; we
list our current Communicative Acts, with descriptive glosses, representative sets of surface patterns,
examples, and other information; we present two labeled dialogues in English and two in Japanese;
and finally, we provide an Appendix describing work in progress.