Proposed Methodology: Combining Knowledge Sources
This methodology falls in the AI Tradition of combining “weak” methods for strong results [Newell, 1973]. It was first proposed by [McRoy, 1992] in the context of Word Sense Disambiguation.
The system combines information from dictionary definitions, selectional restrictions and context subject codes by using a machine learning algorithm to determine the optimal combination of such modules for a given word distribution.
In addition, a part-of-speech tagger is used during a pre-processing step.