Bridging lexical divergence: LLM-assisted, cost-efficient zero-shot scientific entity linking

Scientific domain entity linking (EL) differs from general domain EL because mentions and entity names often lack lexical overlap. Another challenge is that specialized terminology is used in the scientific domain, which is rarely encountered in a model pretrained on a general domain. Therefore, models trained on a general domain transfer poorly to scientific domain. To address these, in-domain fine-tuning is the natural remedy. However, many scientific domains lack expert-annotated data, motivating the need for a zero-shot approach. Existing zero-shot methods heavily rely on LLMs to generate aliases across entire mention corpora, which incurs substantial computational cost, and those methods provide no mechanism to filter out any noise from LLM. To address these challenges, we propose Sci-ZSEL, a zero-shot framework that selectively generates entity aliases with LLM to control computational cost, and applies an ontology-aware filter to remove aliases that semantically drift toward ontology neighbors. Then, filtered aliases are used to construct pseudo-labeled mention-entity pairs for zero-shot fine-tuning. To enable evaluation of EL under low lexical overlap, we also release a new animal science EL benchmark linked to three livestock trait ontologies, where mentions and entities exhibit substantially lower lexical overlap than in existing benchmarks. Across five benchmarks, Sci-ZSEL outperforms the non-fine-tuned baseline, and combining Sci-ZSEL with curated synonyms achieves the best performance in most settings.

Committee: Qi Li (major professor), Ali Jannesari, and Hongyang Gao