Large-Scale Paraphrasing for Natural Language Understanding

We examine the application of data-driven paraphrasing to natural language understanding. We leverage bilingual parallel corpora to extract a large collection of syntactic paraphrase pairs, and introduce an adaptation scheme that allows us to tackle a variety of text transformation tasks via paraphrasing. We evaluate our system on the sentence compression task. Further, we use distributional similarity measures based on context vectors derived from large monolingual corpora to annotate our paraphrases with an orthogonal source of information. This yields significant improvements in our compression system’s output quality, achieving state-of-the-art performance. Finally, we propose a refinement of our paraphrases by classifying them into natural logic entailment relations. By extending the synchronous parsing paradigm towards these entailment relations, we will enable our system to perform recognition of textual entailment.