Learning Containment Metaphors


We present a computational approach that traces the developmental process, from containment image schemas to metaphor, in four phases: a) perceptual discovery of image schemas, b) associating the participants with linguistic units, c) discovering a linguistic structure encoding the schema, and finally d) identifying the distribution of such metaphor usage in a corpus. In the first three phases, we use no prior knowledge about either the perceptual or language domains; in the corpus analysis, we use the WordNet ontology. Our initial image schema, involving a trajector, container and the relation model is correlated with co-occurring commentaries to yield containment-preposition associations through recurring patterns. We then use selectional restrictions to identify the classes of the container words, the most common being location (66%), followed by group membership (20%), time, and cognition (17% each). This process suggests a mechanism whereby language may help refine the initial perceptual image schema.

Back to Table of Contents