Iterative Auto-Annotation for Scientific Named Entity Recognition Using BERT-Based Models

TL;DR

This paper introduces an iterative auto-annotation method using BERT-based models to improve scientific named entity recognition, especially with limited labeled data, by leveraging transfer learning and multiple refinement rounds.

Contribution

The paper proposes a novel iterative fine-tuning approach that enhances SciNER performance using minimal manual annotations and auto-annotated data, outperforming existing models.

Findings

BERT-large-cased outperforms dslim/bert-large-NER in SciNER tasks.

Iterative auto-annotation significantly improves F1 scores.

Method is effective for low-resource labeled data scenarios.

Abstract

This paper presents an iterative approach to performing Scientific Named Entity Recognition (SciNER) using BERT-based models. We leverage transfer learning to fine-tune pretrained models with a small but high-quality set of manually annotated data. The process is iteratively refined by using the fine-tuned model to auto-annotate a larger dataset, followed by additional rounds of fine-tuning. We evaluated two models, dslim/bert-large-NER and bert-largecased, and found that bert-large-cased consistently outperformed the former. Our approach demonstrated significant improvements in prediction accuracy and F1 scores, especially for less common entity classes. Future work could include pertaining with unlabeled data, exploring more powerful encoders like RoBERTa, and expanding the scope of manual annotations. This methodology has broader applications in NLP tasks where access to labeled data is limited.