Data Access for Situation Handling (DASH) 
J.R.Hobbs, Principal Investigator 
Artificial Intelligence Center 
SRI International 
333 Ravenswood Avenue 
Menlo Park CA 94025 
PROJECT GOALS 
The objective of this research is to design a tem- 
plate specification language for representing situations 
and creating data structures for storing information ex- 
tracted from texts. The language will be used to cre- 
ate structured descriptions of those aspects of situations 
about which information is required. These structured 
descriptions will function to 
* characterize the desired information 
• guide the process of extracting information from 
texts 
* provide a representation that can be stored in a 
database for later access 
Experience in developing computer systems to extract 
data from text has shown that determining how to struc- 
ture the information for storage in a database and later 
use can be as significant a problem as configuring a nat- 
urM language system to process texts and extract the 
information. This should not be surprising, since defin- 
ing the representation structures is a special instance of 
the problem of knowledge representation, which is an 
open problem of substantial depth. 
RECENT RESULTS 
The main part of our effort in the first phase of DASH 
has been close analysis of the templates from MUC-4, 
MUC-5, and the Warbreaker Message Handling Tasks. 
We have also attempted to provide templates for the 
a small set of TREC topic descriptions and narratives. 
This analysis has been aimed at explicating general prin- 
ciples of template design that might be presented in a 
manual for template designers. Specifically we have pro- 
posed principles involving the following aspects of tem- 
plate design: 
1. The choice of basic ontology 
2. Issues of temporal granularity 
3. Criteria for choosing among alternative representa- 
tions of relations 
4. How to represent events 
5. Entities vs. Entity Snapshots--time-slices of parts 
of entities 
6. The nature of slot fills 
7. The status of backpointers 
We shall present a report on this research at the Human 
Language Technology Workshop. 
PLANS FOR THE COMING YEAR 
We plan to continue with our specifications of templates 
for TREC topics, with special emphasis on issues involv- 
ing the representation of spatial and temporal relations 
and economic facts. We shall be looking for generaliza- 
tions that might generate further principles of template 
design. We shall be compiling these principles, with ex- 
amples of their application, in a preliminary version of a 
manual for template designers. 
475 
