Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
DGRC AskCal: Natural Language Question Answering for Energy Time Series
 

Summary: DGRC AskCal: Natural Language Question Answering
for Energy Time Series
Andrew Philpot, Jose Luis Ambite, Eduard Hovy
Digital Government Research Center
USC/Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292-6695
{philpot,ambite,hovy}@isi.edu
Abstract
Even quite sophisticated users can experience difficulty navigating large collections of data to locate the answers to
their queries. We describe AskCal, a system that employs natural language processing, an ontology, a query
planner, and various feedback mechanisms to assist a user in refining his or her query and then in executing and
visualizing it. Tracing several interactions with AskCal in the domain of energy time series, we show how a
combination of modalities, including ATN parsing of free-form natural language questions, user modification of
predefined template queries, and fall-back parsing by picking out landmark terms, support a wide variety of user
queries while reducing user query formulation effort. We illustrate the use of feedback mechanisms to guide the
user toward regions of the query space where useful data can be found.
1. Introduction
Over that past few years, the Digital Government Research Center1
's Energy Data Collection (EDC) project has

  

Source: Ambite, Josť Luis - Information Sciences Institute & Department of Computer Science, University of Southern California
Hovy, Eduard - Information Sciences Institute, University of Southern California

 

Collections: Computer Technologies and Information Sciences