The Perils and Rewards of Developing                                        
Restricted Domain Applications 
Daniel Marcu 
Information Sciences Institute and 
Department of Computer Science 
University of Southern California 
4676 Admiralty Way, Suite 1001 
Marina del Rey, CA 90292 
marcu@isi.edu 
Abstract 
Over the last three decades, the development of 
restricted domain applications has been an ongo-
ing theme in computational linguistic research. 
Speech Recognition, Machine Translation, Sum-
marization, and Question Answering researchers 
have all built at one time or another restricted 
domain systems. In retrospect, it is long due to 
examine both the successes and failures of these 
previous attempts. 
In this talk, I will examine the circumstances 
in which the development of restricted domain 
applications has led to significant advances in the 
state of the art and the circumstances in which 
restricted domain research has had little impact 
on our field. I will use the lessons learned from 
previous attempts to building restricted domain 
natural language processing applications in order 
to examine the potential impact of current re-
search in restricted domain question answering. 
  
   
