Benchmarking for syntax-based sentential inference

摘要

We propose a methodology for investigat-ing how well NLP systems handle mean-ing preserving syntactic variations. We start by presenting a method for the semi automated creation of a benchmark where entailment is mediated solely by meaning preserving syntactic variations. We then use this benchmark to compare a seman-tic role labeller and two grammar based RTE systems. We argue that the proposed methodology (i) supports a modular eval-uation of the ability of NLP systems to handle the syntax/semantic interface and (ii) permits focused error mining and er-ror analysis.