On-the-fly Definition Augmentation of LLMs for Biomedical NER

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Munnangi, Monica, Feldman, Sergey, Wallace, Byron C, Amir, Silvio, Hope, Tom, Naik, Aakanksha
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929324998787072
author Munnangi, Monica
Feldman, Sergey
Wallace, Byron C
Amir, Silvio
Hope, Tom
Naik, Aakanksha
author_facet Munnangi, Monica
Feldman, Sergey
Wallace, Byron C
Amir, Silvio
Hope, Tom
Naik, Aakanksha
contents Despite their general capabilities, LLMs still struggle on biomedical NER tasks, which are difficult due to the presence of specialized terminology and lack of training data. In this work we set out to improve LLM performance on biomedical NER in limited data settings via a new knowledge augmentation approach which incorporates definitions of relevant concepts on-the-fly. During this process, to provide a test bed for knowledge augmentation, we perform a comprehensive exploration of prompting strategies. Our experiments show that definition augmentation is useful for both open source and closed LLMs. For example, it leads to a relative improvement of 15\% (on average) in GPT-4 performance (F1) across all (six) of our test datasets. We conduct extensive ablations and analyses to demonstrate that our performance improvements stem from adding relevant definitional knowledge. We find that careful prompting strategies also improve LLM performance, allowing them to outperform fine-tuned language models in few-shot settings. To facilitate future research in this direction, we release our code at https://github.com/allenai/beacon.
format Preprint
id arxiv_https___arxiv_org_abs_2404_00152
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On-the-fly Definition Augmentation of LLMs for Biomedical NER
Munnangi, Monica
Feldman, Sergey
Wallace, Byron C
Amir, Silvio
Hope, Tom
Naik, Aakanksha
Computation and Language
Despite their general capabilities, LLMs still struggle on biomedical NER tasks, which are difficult due to the presence of specialized terminology and lack of training data. In this work we set out to improve LLM performance on biomedical NER in limited data settings via a new knowledge augmentation approach which incorporates definitions of relevant concepts on-the-fly. During this process, to provide a test bed for knowledge augmentation, we perform a comprehensive exploration of prompting strategies. Our experiments show that definition augmentation is useful for both open source and closed LLMs. For example, it leads to a relative improvement of 15\% (on average) in GPT-4 performance (F1) across all (six) of our test datasets. We conduct extensive ablations and analyses to demonstrate that our performance improvements stem from adding relevant definitional knowledge. We find that careful prompting strategies also improve LLM performance, allowing them to outperform fine-tuned language models in few-shot settings. To facilitate future research in this direction, we release our code at https://github.com/allenai/beacon.
title On-the-fly Definition Augmentation of LLMs for Biomedical NER
topic Computation and Language
url https://arxiv.org/abs/2404.00152