[Colloquium] Today: Boven/MS Presentation/May 14, 2007

[Margaret Jaffey]

This is a reminder about Bradley Boven's MS Presentation this afternoon.

----------------------------
Date:  Monday, May 14, 2007

Time:  3:30 p.m.

Place:  Ryerson 276

M.S. Candidate:  Bradley Boven

M.S. Paper Title:  Approaches and Issues in Dialog Act Tagging

Abstract:
Dialog acts are a key component of language that provide a basis for  
spoken language understanding between humans.  A dialog act (DA)  
represents the meaning of an utterance at the level of an  
illocutionary force [Stolcke et al.,2000].  They convey not the  
semantic meaning of an utterance, but rather the function an  
utterance is intended to serve.  Examples of these labels include  
questions, statements, backchannels, floor controls, and disruptions.

Dialog acts are an important aspect of language as they help both  
humans and language systems understand what a user is trying to  
convey semantically by first understanding their words functionally.   
Thus, predicting dialog act tags has been studied for several decades  
as a means to improve our language understanding systems as a whole.  
In this paper we analyze many of the influential DA tagging  
approaches to date. In doing this, we explain many of the influential  
corpora and DA tagsets, as well as the tagging schemes themselves,  
breaking down both the techniques used and the feature sets that make  
up each model.  We also try to understand the driving motivation  
behind each different strategy to help the reader understand why the  
feature sets and their combined techniques were chosen. We then  
investigate these approaches as a collection to determine what  
features are most important to creating an accurate dialog act  
tagger, and also what techniques work best when paired with these  
different feature sets.  We then investigate new research aimed at  
quantitatively investigating which of these feature sets work best  
both individually and in combination with one another, when applied  
to different tagging techniques and classifiers.   Finally, we look  
at what has approaches have been successful and what challenges we  
face in the future as we try to design new algorithms to tackle this  
complex research question.

Advisors: Professors Gina-Anne Levow and John Goldsmith

A draft copy of Brad Boven's MS Paper is available in Ry 161A.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Margaret P. Jaffey                             margaret at cs.uchicago.edu
Department of Computer Science
Student Support Rep (Ry 161A)        (773) 702-6011
The University of Chicago                  http://www.cs.uchicago.edu
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=