Title: ChatGPT to identify drug-drug interactions from texts
Presenter: Hasin Rehana (Univ. of North Dakota; USA)
Abstract
Drug-drug interactions (DDIs) denote changes in a drug's effect when co-administered with another drug. Such interactions can either diminish or enhance the therapeutic effects of one or both drugs, potentially leading to adverse side effects. Adverse drug reactions account for approximately 0.3 million deaths annually in the United States and Europe. Consequently, accurately diagnosing DDIs is crucial for public health safety. With the burgeoning volume of biomedical literature, there is a heightened focus on developing automated methods for DDI extraction. This study evaluates the performance of Large Language Models - GPT-3, GPT-3.5, and GPT-4, different iterations of ChatGPT, using a subset of the DDI 2013 dataset. Our findings indicate that the latest model, GPT-4, achieved an impressive accuracy of 97.88%, setting a new benchmark. These promising outcomes underscore the growing significance and potential of large language models in biomedical text analysis and DDI identification.
Biomedical Ontology World is a channel to publish talks, lessons, and interviews for ontology development and implementation for bio and health big data, data interoperability, AI, and knowledge presentation.
Негізгі бет Ғылым және технология VDOS2023 Talk#6 Hasin - ChatGPT to identify drug-drug interactions from texts
Пікірлер