NANER: Instance-Aware Named Entity Recognition (NER)
- Proposed NANER, an NER model that utilized instance-based prompt learning to resolve issues related to category ambiguity and the complexity of obtaining high-quality descriptions.
- Employed an instance-based span model named NASpan within NANER, constructing spans with complete tokens, guided by specific entity instances sampled from training sets or online sources like Wikipedia.
- Verified the robustness of NANER by achieving state-of-the-art F1 improvements on datasets such as ACE04, ACE05, and GENIA. NANER also excelled in domain transfer tasks through zero & few-shot learning, enhancing F1 of 11.13% on CoNLL03 and 8.59% on Wnut17 compared to description-based zero-shot benchmarks.
- Conclusion: In this paper, we proposed a prompt learning method and designed an instance-based span model for named entity recognition. We sample instances from training sets or online sources and use instances for enhanced span construction which contains complete span information. Experimental results and empirical analysis demonstrate that NANER achieves SOTA performance under both supervised and domain transfer settings, which verifies its effectiveness and generalization ability. Besides, based on instance learning, we proposed a fast and effective ensemble method called instance ensemble, which saves many time and computation resources compared with the traditional model ensemble. In future work, we plan to explore how to automatically obtain instances, identify entities in parallel, and extend instance learning to handle more NLP downstream tasks.
Zelin Li
My research interests include Natural Language Processing, Large Language Models and Data Science.