Back to Search
Start Over
Improving BERT with Self-Supervised Attention
- Publication Year :
- 2020
-
Abstract
- One of the most popular paradigms of applying large pre-trained NLP models such as BERT is to fine-tune it on a smaller dataset. However, one challenge remains as the fine-tuned model often overfits on smaller datasets. A symptom of this phenomenon is that irrelevant or misleading words in the sentence, which are easy to understand for human beings, can substantially degrade the performance of these finetuned BERT models. In this paper, we propose a novel technique, called Self-Supervised Attention (SSA) to help facilitate this generalization challenge. Specifically, SSA automatically generates weak, token-level attention labels iteratively by probing the fine-tuned model from the previous iteration. We investigate two different ways of integrating SSA into BERT and propose a hybrid approach to combine their benefits. Empirically, through a variety of public datasets, we illustrate significant performance improvement using our SSA-enhanced BERT model.
- Subjects :
- Computer Science - Computation and Language
Computer Science - Machine Learning
Subjects
Details
- Database :
- arXiv
- Publication Type :
- Report
- Accession number :
- edsarx.2004.03808
- Document Type :
- Working Paper
- Full Text :
- https://doi.org/10.1109/ACCESS.2021.3122273