Preposition Semantic Classification via Treebank and FrameNet

This paper reports on experiments in classifying the semantic role annotations assigned to prepositional phrases in both PENN TREEBANK (version II) and FRAMENET (version 0.75). In both cases, experiments are done to see how the prepositions can be classified given the dataset’s role inventory, using standard word-sense disambiguation features, such as the parts of speech of surrounding words, and collocations indicative of the particular roles. In addition to using traditional word collocations, the experiments incorporate class-based collocations in the form of WordNet hypernyms. Separate classifiers are produced for each preposition. For TreeBank, the wordcollocations achieve slightly better performance: 78.5% versus 77.4%. However, for FrameNet, the combined collocations achieve better performance: 70.3% versus 68.5% Furthermore when classifying all the TreeBanks prepositions together, the combined yields a noticable gain at 85.8% accuracy versus 81.3% for word-only collocations.