Large Scale Acquisition of Paraphrases for Learning Surface Patterns

Paraphrases have proved to be useful in many applications, including Machine Translation, Question Answering, Summarization, and Information Retrieval. Paraphrase acquisition methods that use a single monolingual corpus often produce only syntactic paraphrases. We present a method for obtaining surface paraphrases, using a 150GB (25 billion words) monolingual corpus. Our method achieves an accuracy of around 70% on the paraphrase acquisition task. We further show that we can use these paraphrases to generate surface patterns for relation extraction. Our patterns are much more precise than those obtained by using a state of the art baseline and can extract relations with more than 80% precision for each of the test relations.