Chem-FINESE: Validating Fine-Grained Few-shot Entity Extraction through Text Reconstruction

• DOI: 10.48550/arxiv.2401.10189
• Publication Date: 2024-01-01
• Authors: Qingyun Wang; Zixuan Zhang; Hongxiang Li; Xuan Liu; Jiawei Han; Heng Ji; Huimin Zhao
• Keywords: Extractor; Schema (genetic algorithms); Entity linking; Sequence labeling; Sequence (biology)

Fine-grained few-shot entity extraction in the chemical domain faces two unique challenges. First, compared with tasks general domain, sentences from papers usually contain more entities. Moreover, models have difficulty extracting entities of long-tailed types. In this paper, we propose Chem-FINESE, a novel sequence-to-sequence (seq2seq) based approach, to address these Our Chem-FINESE has components: seq2seq extractor extract named input sentence and self-validation module reconstruct original extracted Inspired by fact that good system needs faithfully, our new leverages results sentence. Besides, design contrastive loss reduce excessive copying during process. Finally, release ChemNER+, fine-grained dataset is annotated experts ChemNER schema. Experiments settings both ChemNER+ CHEMET datasets show newly proposed framework contributed up 8.26% 6.84% absolute F1-score gains respectively.