Selecting on-topic sentences from natural language corpora
Michael Levit, Elizabeth Boschee, Marjorie R. Freedman · 2007
We describe a system that examines input sentences with re-spect to arbitrary topics formulated as natural language expres-sions. It extracts predicate-argument structures from text in-tervals and links them into semantically organized proposition trees. By instantiating trees constructed for topic descriptions in trees representing input sentences or parts thereof, we are able to assess degree of “topicality ” for each sentence. The presented strategy was used in the BBN distillation system for the GALE Year 1 evaluation and achieved outstanding results compared to other systems and human participants. Index Terms: machine learning, question answering, topicality. 1.