|
| Speaker: | Owen Rambow |
| CoGenTex Co., Ithaca, USA | |
| Date: | Wednesday 8th September 1999 |
| Time: | 14:00-15:30 |
| Place: | E6A211 (JRCASE Seminar Room), Macquarie University |
Abstract: For planning extended discourse, (at least) two types of approaches have been proposed in the literature. In the first approach, general knowledge about discourse and its structure is used to reason about the speaker's (S's) communicative goals in order to assemble a tree representing the content and structure of the discourse, which, when delivered to the hearer (H), will meet S's goals. We will call this the ``rhetorical approach''. In a second approach, the text structure is directly specified using a simple device, no intentions are modeled, and no reasoning takes place. We will call this the ``schema approach''.
While at first one may assume that the schema approach is just a ``compilation'' (in some sense to be made specific) of the rhetorical approach, we argued in Kittredge et al (1991) that certain kinds of linguistic communications -- namely, reports -- cannot be adequately planned using general knowledge about discourse structure, because their structure has been, through historical precedent, fixed in some now arbitrary manner (even though historically, it may have been motivated). Thus, while the rhetorical approach is theoretically interesting and has explanatory power, it is empirically inadequate.
In this talk, I propose a way of integrating the two approaches into a single, theoretically interesting and empirically adequate framework. This framework is based on the notion of {\em extended discourse plan operators}. I propose that in planning discourse, the planning agent assembles large pieces of structure on several levels of representation in parallel. The proposed approach also solves another problem: if, as has been argued by Moore and Pollack (1992), discourse can be represented by parallel levels of structure which are not necessarily isomorphic, how can we plan such parallel structure?
The analysis of existing approaches is similar to that provided Moore and Paris (1993) and of Moore (1994). Like those authors, this paper proposes a way of integrating different types of discourse-related information in a planning system. However, the approach chosen is different.
About the Presenter: Owen Rambow obtained his Ph.D. in computational linguistics from the University of Pennsylvania in 1994, and has worked on automatic text generation technology since 1987, participating in the design and implementation of several text generation systems. He has also worked in natural language understanding, as well as formal and mathematical linguistics.
Dr. Rambow was a co-founder of CoGenTex (Ithaca) with Dr. Richard Kittredge and Dr. Tanya Korelsky in 1990. Since that time he has conducted research on theoretical issues in syntax as well as applications. His research interests include the notion of syntactic dependency, formal models of syntax, computational models of parsing, and the proper linguistic representation of word order. His work on text generation has concentrated on applications in the software engineering domain and the role of domain-specific and sublanguage information.
Dr. Rambow has been responsible for CoGenTex's cooperation with the University of Pennsylvania, aimed at transitioning parsing technology developed there into robust application systems. He is also project leader for the machine translation project, as well as for the KACS project.
Enquiries: sals@mri.mq.edu.au
| Last modified: August, 1999 |