We describe two experiments related to learning new words in context. We study two types of new words: metaphors (for whom a related meaning already exists) and artificial words. The new words were used anaphorically to refer to past objects in the text. For anaphoric metaphors, subjects showed an initial bias to adopt a literal interpretation, which shifted as the experiment progressed. Subjects learned the meaning of the metaphors more rapidly and more accurately. After repeated exposure to the words in appropriate contexts, metaphoric sentences were processed comparably with the sentences made only of familiar words, whereas artificial-word sentences maintained a slight disadvantage. Results suggest that participants used context matching to understand and learn new words. We present a computational model that captures the essential trends in the data obtained from the two experiments.

41 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu