Inferring parts of speech for lexical mappings via the Cyc KB

Inferring parts of speech for lexical mappings via the Cyc KB Tom O’Hara‡, Stefano Bertolo, Michael Witbrock, Bjørn Aldag, Jon Curtis, with Kathy Panton, Dave Schneider, and Nancy Salay ‡Computer Science Department New Mexico State University Las Cruces, NM 88001 tomohara@cs.nmsu.edu Cycorp, Inc. Austin, TX 78731 {bertolo,witbrock,aldag}@cyc.com {jonc,panton,daves,nancy}@cyc.com Abstract We present an automatic approach to learn- ing criteria for classifying the parts-of-speech used in lexical mappings. This will fur- ther automate our knowledge acquisition sys- tem for non-technical users. The criteria for the speech parts are based on the types of the denoted terms along with morphological and corpus-based clues. Associations among these and the parts-of-speech are learned us- ing the lexical mappings contained in the Cyc knowledge base as training data. With over 30 speech parts to choose from, the classifier achieves good results (77.8% correct). Ac- curate results (93.0%) are achieved in the special case of the mass-count distinction for nouns. Comparable results are also obtained using OpenCyc (73.1% general and 88.4% mass-count). 1 Introduction In semantic lexicons, the term lexical mapping de- scribes the relation between a concept and a phrase used to refer to it (Onyshkevych and Nirenburg, 1995; Burns and Davis, 1999). Lexical mappings include associated syntactic information, in partic- ular, part of speech information for phrase head- words. The term lexicalize will refer to the process of producing these mappings, which are referred to as lexicalizations. Selecting the part of speech for the lexical mapping is required so that proper inflectional variations can be recognized and gen- erated for the term. Although producing lexi- calizations is often a straightforward task, there are many cases that can pose problems, especially when fine-grained speech part categories are used. For example, the headword ‘painting’ is a verb in the phrase “painting for money” but a noun in the phrase “painting for sale.” In cases like this, semantic or pragmatic criteria, as opposed to syntactic criteria only, are necessary for deter- mining the proper part of speech. The headword part of speech is important for correctly identifying phrasal variations. For instance, the first term can also occur in the same sense in “paint for money.” However, this does not hold for the second case, since “paint for sale” has an entirely different sense (i.e., a substance rather than an artifact). When lexical mappings are produced by naive users, such as in DARPA’s Rapid Knowledge For- mation (RKF) project, it is desirable that tech- nical details such as the headword part of speech be inferred for the user. Otherwise, often complex and time-consuming clarification dialogs might be necessary in order to rule out various possibilities. For example, Cycorp’s Dictionary Assistant was developed for RKF in order to allow non-technical users to specify lexical mappings from terms into the Cyc knowledge base (KB). Currently, when a new type of activity is described, the user is asked a series of questions about the ways of referring to the activity. If the user enters the phrase “painting for money,” the system asks …

Content truncated to 500 words