SPECTER: Document-level Representation Learning using Citation-informed Transformers

Arman Cohan, Sergey Feldman, Iz Beltagy, Doug Downey, Daniel Weld


Abstract
Representation learning is a critical ingredient for natural language processing systems. Recent Transformer language models like BERT learn powerful textual representations, but these models are targeted towards token- and sentence-level training objectives and do not leverage information on inter-document relatedness, which limits their document-level representation power. For applications on scientific documents, such as classification and recommendation, accurate embeddings of documents are a necessity. We propose SPECTER, a new method to generate document-level embedding of scientific papers based on pretraining a Transformer language model on a powerful signal of document-level relatedness: the citation graph. Unlike existing pretrained language models, Specter can be easily applied to downstream applications without task-specific fine-tuning. Additionally, to encourage further research on document-level models, we introduce SciDocs, a new evaluation benchmark consisting of seven document-level tasks ranging from citation prediction, to document classification and recommendation. We show that Specter outperforms a variety of competitive baselines on the benchmark.
Anthology ID:
2020.acl-main.207
Volume:
Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
Month:
July
Year:
2020
Address:
Online
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
2270–2282
Language:
URL:
https://www.aclweb.org/anthology/2020.acl-main.207
DOI:
10.18653/v1/2020.acl-main.207
Bib Export formats:
BibTeX MODS XML EndNote
PDF:
http://aclanthology.lst.uni-saarland.de/2020.acl-main.207.pdf
Video:
 http://slideslive.com/38928763