A Bootstrapping Method for Learning Semantic Lexicons using Extraction Pattern Contexts

A Bootstrapping Method for Learning Semantic Lexicons using Extraction Pattern Contexts Michael Thelen and Ellen Riloff School of Computing University of Utah Salt Lake City, UT 84112 USA {thelenm,riloff}@cs.utah.edu Abstract This paper describes a bootstrapping al- gorithm called Basilisk that learns high- quality semantic lexicons for multiple cate- gories. Basilisk begins with an unannotated corpus and seed words for each semantic category, which are then bootstrapped to learn new words for each category. Basilisk hypothesizes the semantic class of a word based on collective information over a large body of extraction pattern contexts. We evaluate Basilisk on six semantic categories. The semantic lexicons produced by Basilisk have higher precision than those produced by previous techniques, with several cate- gories showing substantial improvement. 1 Introduction In recent years, several algorithms have been devel- oped to acquire semantic lexicons automatically or semi-automatically using corpus-based techniques. For our purposes, the term semantic lexicon will refer to a dictionary of words labeled with semantic classes (e.g., “bird” is an animal and “truck” is a vehicle). Semantic class information has proven to be useful for many natural language processing tasks, includ- ing information extraction (Riloff and Schmelzen- bach, 1998; Soderland et al., 1995), anaphora resolu- tion (Aone and Bennett, 1996), question answering (Moldovan et al., 1999; Hirschman et al., 1999), and prepositional phrase attachment (Brill and Resnik, 1994). Although some semantic dictionaries do exist (e.g., WordNet (Miller, 1990)), these resources often do not contain the specialized vocabulary and jargon that is needed for specific domains. Even for rela- tively general texts, such as the Wall Street Journal (Marcus et al., 1993) or terrorism articles (MUC- 4 Proceedings, 1992), Roark and Charniak (Roark and Charniak, 1998) reported that 3 of every 5 terms generated by their semantic lexicon learner were not present in WordNet. These results suggest that auto- matic semantic lexicon acquisition could be used to enhance existing resources such as WordNet, or to produce semantic lexicons for specialized domains. We have developed a weakly supervised bootstrap- ping algorithm called Basilisk that automatically generates semantic lexicons. Basilisk hypothesizes the semantic class of a word by gathering collective evidence about semantic associations from extraction pattern contexts. Basilisk also learns multiple se- mantic classes simultaneously, which helps constrain the bootstrapping process. First, we present Basilisk’s bootstrapping algo- rithm and explain how it differs from previous work on semantic lexicon induction. Second, we present empirical results showing that Basilisk outperforms a previous algorithm. Third, we explore the idea of learning multiple semantic categories simultaneously by adding this capability to Basilisk as well as an- other bootstrapping algorithm. Finally, we present results showing that learning multiple semantic cat- egories simultaneously improves performance. 2 Bootstrapping using Collective Evidence from Extraction Patterns Basilisk (Bootstrapping Approach to SemantIc Lexicon Induction using Semantic Knowledge) is a weakly supervised bootstrapping algorithm that au- tomatically generates semantic lexicons. Figure 1 shows the high-level view of Basilisk’s bootstrapping process. The input to Basilisk is an unannotated text corpus and a few manually defined seed words for …

Content truncated to 500 words