Automated dictionary construction for information extraction from text

Abstract

The authors have developed a tool called AutoSlog that automatically constructs domain-specific dictionaries given a set of annotated training texts. Using AutoSlog, a first-year graduate student who had minimal experience with the CIRCUS sentence analyzer on which AutoSlog is based, created a dictionary for the domain of terrorism in 8 hours. In the experiments, the 8-hour AutoSlog dictionary achieved 90% of the performance of a hand-crafted dictionary that required 1500 person-hours of effort by 2 advanced graduate students who were highly skilled with the sentence analyzer.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>