Robust Textual Inference using Diverse Knowledge Sources

We present a machine learning approach to ro- bust textual inference, in which parses of the text and the hypothesis sentences are used to mea- sure their asymmetric "similarity", and thereby to decide if the hypothesis can be inferred. This idea is realized in two different ways. In the first, each sentence is represented as a graph (extracted from a dependency parser) in which the nodes are words/phrases, and the links represent depen- dencies. A learned, asymmetric, graph-matching cost is then computed to measure the similar- ity between the text and the hypothesis. In the second approach, the text and the hypothesis are parsed into the logical formula-like representa- tion used by (Harabagiu et al., 2000). An abduc- tive theorem prover (using learned costs for mak- ing different types of assumptions in the proof) is then applied to try to infer the hypothesis from the text, and the total "cost" of proving the hy- pothesis is used to decide if the hypothesis is en- tailed.