Learning to Selectively Learn for Weakly-supervised Paraphrase Generation

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 5930–5940 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 5930 Learning to Selectively Learn for Weakly-supervised Paraphrase Generation Kaize Ding1∗ Dingcheng Li2 Alexander Hanbo Li3 Xing Fan2 Chenlei Guo2 Yang Liu2 Huan Liu1 1Arizona State University 2Amazon Alexa AI 3Amazon AWS AI {kaize.ding,huan.liu}@asu.edu {lidingch,hanboli,fanxing,guochenl,yangliud}@amazon.com Abstract Paraphrase generation is a longstanding NLP task that has diverse applications for down- stream NLP tasks. However, the effective- ness of existing efforts predominantly relies on large amounts of golden labeled data. Though unsupervised endeavors have been proposed to address this issue, they may fail to generate meaningful paraphrases due to the lack of su- pervision signals. In this work, we go beyond the existing paradigms and propose a novel approach to generate high-quality paraphrases with weak supervision data. Specifically, we tackle the weakly-supervised paraphrase gen- eration problem by: (1) obtaining abundant weakly-labeled parallel sentences via retrieval- based pseudo paraphrase expansion; and (2) developing a meta-learning framework to pro- gressively select valuable samples for fine- tuning a pre-trained language model, i.e., BART, on the sentential paraphrasing task. We demonstrate that our approach achieves signifi- cant improvements over existing unsupervised approaches, and is even comparable in perfor- mance with supervised state-of-the-arts. 1 Introduction Paraphrase generation is a fundamental NLP task that restates text input in a different surface form while preserving its semantic meaning. It serves as a cornerstone in a wide spectrum of NLP appli- cations, such as question answering (Dong et al., 2017), machine translation (Resnik et al., 2010), and semantic parsing (Berant and Liang, 2014). With the recent advances of neural sequence-to- sequence (Seq2Seq) architecture in the field of lan- guage generation, a growing amount of literature has also applied Seq2Seq models to the sentential paraphrasing task. Despite their promising results, collecting large amounts of parallel paraphrases is often time- consuming and requires intensive domain knowl- ∗Work was done as an intern at Amazon Alexa AI. edge. Therefore, the performance of supervised methods could be largely limited in real-world sce- narios. Due to this problem, unsupervised para- phrase generation has recently received increasing attention, but the development is still in its infancy. Generally, sampling-based or editing-based ap- proaches (Bowman et al., 2016; Miao et al., 2019) fail to incorporate valuable supervised knowledge, resulting in less coherent and controllable gener- ated paraphrases (Liu et al., 2019). In this work, we propose going beyond the existing learning paradigms and investigate a novel research prob- lem – weakly-supervised paraphrase generation, in order to push forward the performance boundary of sentential paraphrasing models with low-cost supervision signals. As an understudied problem, weakly-supervised paraphrase generation is challenging mainly be- cause of the following reasons: (i) although weak supervision has been applied in different low- resource NLP tasks (Dehghani et al., 2017; Aprosio et al., 2019), for paraphrase generation, it is unclear how to automatically acquire abundant weak super- vision data that contains coherent, fluent and di- verse paraphrases; (ii) weakly-labeled paraphrases tend to be noisy …

Content truncated to 500 words