Enhancing Large Language Models with Domain-specific Retrieval Augment Generation: A Case Study on Long-form Consumer Health Question Answering in Ophthalmology

Despite the potential of Large Language Models (LLMs) in medicine, they may generate responses lacking supporting evidence or based on hallucinated evidence. While Retrieval Augment Generation (RAG) is popular to address this issue, few studies implemented and evaluated RAG in downstream domain-specific applications. We developed a RAG pipeline with 70,000 ophthalmology-specific documents that retrieve relevant documents to augment LLMs during inference time. In a case study on long-form consumer health questions, we systematically evaluated the responses including over 500 references of LLMs with and without RAG on 100 questions with 10 healthcare professionals. The evaluation focuses on factuality of evidence, selection and ranking of evidence, attribution of evidence, and answer accuracy and completeness. LLMs without RAG provided 252 references in total. Of which, 45.3% hallucinated, 34.1% consisted of minor errors, and 20.6% were correct. In contrast, LLMs with RAG significantly improved accuracy (54.5% being correct) and reduced error rates (18.8% with minor hallucinations and 26.7% with errors). 62.5% of the top 10 documents retrieved by RAG were selected as the top references in the LLM response, with an average ranking of 4.9. The use of RAG also improved evidence attribution (increasing from 1.85 to 2.49 on a 5-point scale, P<0.001), albeit with slight decreases in accuracy (from 3.52 to 3.23, P=0.03) and completeness (from 3.47 to 3.27, P=0.17). The results demonstrate that LLMs frequently exhibited hallucinated and erroneous evidence in the responses, raising concerns for downstream applications in the medical domain. RAG substantially reduced the proportion of such evidence but encountered challenges.

Comparison of pipeline, sequence-to-sequence, and GPT models for end-to-end relation extraction: experiments with the rare disease use-case

End-to-end relation extraction (E2ERE) is an important and realistic application of natural language processing (NLP) in biomedicine. In this paper, we aim to compare three prevailing paradigms for E2ERE using a complex dataset focused on rare diseases involving discontinuous and nested entities. We use the RareDis information extraction dataset to evaluate three competing approaches (for E2ERE): NER $\rightarrow$ RE pipelines, joint sequence to sequence models, and generative pre-trained transformer (GPT) models. We use comparable state-of-the-art models and best practices for each of these approaches and conduct error analyses to assess their failure modes. Our findings reveal that pipeline models are still the best, while sequence-to-sequence models are not far behind; GPT models with eight times as many parameters are worse than even sequence-to-sequence models and lose to pipeline models by over 10 F1 points. Partial matches and discontinuous entities caused many NER errors contributing to lower overall E2E performances. We also verify these findings on a second E2ERE dataset for chemical-protein interactions. Although generative LM-based methods are more suitable for zero-shot settings, when training data is available, our results show that it is better to work with more conventional models trained and tailored for E2ERE. More innovative methods are needed to marry the best of the both worlds from smaller encoder-decoder pipeline models and the larger GPT models to improve E2ERE. As of now, we see that well designed pipeline models offer substantial performance gains at a lower cost and carbon footprint for E2ERE. Our contribution is also the first to conduct E2ERE for the RareDis dataset.

End-to-End Models for Chemical-Protein Interaction Extraction: Better Tokenization and Span-Based Pipeline Strategies

End-to-end relation extraction (E2ERE) is an important task in information extraction, more so for biomedicine as scientific literature continues to grow exponentially. E2ERE typically involves identifying entities (or named entity recognition (NER)) and associated relations, while most RE tasks simply assume that the entities are provided upfront and end up performing relation classification. E2ERE is inherently more difficult than RE alone given the potential snowball effect of errors from NER leading to more errors in RE. A complex dataset in biomedical E2ERE is the ChemProt dataset (BioCreative VI, 2017) that identifies relations between chemical compounds and genes/proteins in scientific literature. ChemProt is included in all recent biomedical natural language processing benchmarks including BLUE, BLURB, and BigBio. However, its treatment in these benchmarks and in other separate efforts is typically not end-to-end, with few exceptions. In this effort, we employ a span-based pipeline approach to produce a new state-of-the-art E2ERE performance on the ChemProt dataset, resulting in $> 4\%$ improvement in F1-score over the prior best effort. Our results indicate that a straightforward fine-grained tokenization scheme helps span-based approaches excel in E2ERE, especially with regards to handling complex named entities. Our error analysis also identifies a few key failure modes in E2ERE for ChemProt.