Disambiguating Open Formed Compound Nouns in Natural Language Processing

Abstract

Natural Language Processing is the field of Computer Science that tries to derive meaning from human languages while also trying to mimic them. One large problem in the field of linguistics is compound noun disambiguation. Words like living room and swimming pool are examples of compound nouns we use everyday and whose meanings are very well understood by humans, yet computers have a hard time interpreting these words. Compound nouns can be created by placing almost any two nouns together, thus creating an open-ended process that allows many different semantic relationships to exist between the nouns; even if the machine knows the meaning of the individual nouns, it is not enough to predict the semantic relationship between them. For example, when comparing steel knife and butter knife, one is a knife made OF steel while the other is a knife FOR cutting butter. Both compounds are used to modify the word knife but they have different semantic relationships with the modifier noun. My program tries to classify the semantic relationships of compound nouns by using WordNet, a lexical database, to find words that have a similar meaning in a hand created corpus of classified compound nouns. I am able to correctly classify the compound nouns 50% of the time, which performs as well, if not better, than other state of the art techniques.