Textual Relatedness based on Predicate-Argument Structure
Speaker: Rui Wang
Abstract:
In this presentation, I would like to talk about my recent work on recognizing textual entailment. I will give some background knowledge of the problem and then define a new measurement for one pair of texts, called Textual Relatedness, which is a weaker concept than semantic similarity or paraphrase. Then I will show that an alignment model based on the predicate-argument structure can effectively capture this relationship and further help an RTE system to classify the Unknown cases at the first stage and thus improve the overall performance as well. In addition, several heterogeneous lexical resources are tested in order to show the different contributions from them.