SpanBERT: Improving Pre-training by Representing and Predicting Spans

• Published in: Transactions of the Association for Computational Linguistics Vol. 8; pp. 64 - 77
• Main Authors: Joshi, Mandar, Chen, Danqi, Liu, Yinhan, Weld, Daniel S., Zettlemoyer, Luke, Levy, Omer
• Format: Journal Article
• Language: English
• Published: One Rogers Street, Cambridge, MA 02142-1209, USA MIT Press 01.01.2020; MIT Press Journals, The; The MIT Press
• Subjects: Artificial intelligence; Boundary representation; Computational linguistics; Computer science; Linguistics; Masking; Training
• ISSN: 2307-387X; 2307-387X
• DOI: 10.1162/tacl_a_00300

We present SpanBERT, a pre-training method that is designed to better represent and predict spans of text. Our approach extends BERT by (1) masking contiguous random spans, rather than random tokens, and (2) training the span boundary representations to predict the entire content of the masked span, without relying on the individual token representations within it. SpanBERT consistently outperforms BERT and our better-tuned baselines, with substantial gains on span selection tasks such as question answering and coreference resolution. In particular, with the same training data and model size as BERT , our single model obtains 94.6% and 88.7% F1 on SQuAD 1.1 and 2.0 respectively. We also achieve a new state of the art on the OntoNotes coreference resolution task (79.6% F1), strong performance on the TACRED relation extraction benchmark, and even gains on GLUE.