Extracting and filtering paraphrases by bridging natural language inference and paraphrasing

Abstract

Paraphrasing is a useful natural language processing task that can contribute to more diverse generated or translated texts. Natural language inference (NLI) and paraphrasing share some similarities and can benefit from a joint approach. We propose a novel methodology for the extraction of paraphrasing datasets from NLI datasets and cleaning existing paraphrasing datasets. Our approach is based on bidirectional entailment; namely, if two sentences can be mutually entailed, they are paraphrases. We evaluate our approach using several large pretrained transformer language models in the monolingual and cross-lingual setting. The results show high quality of extracted paraphrasing datasets and surprisingly high noise levels in two existing paraphrasing datasets.

References in corpus (6)

• Decoupled Weight Decay Regularization
• Microsoft COCO Captions: Data Collection and Evaluation Server
• XNLI: Evaluating Cross-lingual Sentence Representations
• Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks
• Open Subtitles Paraphrase Corpus for Six Languages
• Can NLI Models Verify QA Systems' Predictions?