Yüksek LisansAçık Erişim

Using contrastive learning and large language models on biomedical information extraction

Bu tez size mi ait?

Bu kayıt toplu arşivden geldi. Sizinse profilinize bağlayın.

2024
0 görüntülenme
0 i̇ndirme

Özet (EN)

As the volume of human knowledge grows at an extraordinary rate, it is becoming progressively challenging to fully comprehend and utilize the vast array of information available, even within a specialized domain. The field of Biomedicine exemplifies this, with thousands of research papers published each year contributing new insights about species, diseases, and chemicals. Such growth has created an overwhelming need for effective Information Retrieval systems, coupled with Natural Language Processing techniques, to enable the extraction and use of relevant data at a scale beyond human capacity. Effective Information Retrieval in Biomedicine demands an understanding of published research at a granular level, which involves solving several interconnected challenges. For each Biomedical abstract, it is necessary to recognize entities, normalize these entities with standardized identifiers, and extract relationships between different types of entities. These tasks facilitate a structured representation of knowledge and make complex Biomedical information more accessible for scientific and clinical appli- cations. In this study, we investigate and enhance the performance of an existing normal- ization tool, BioNEN [1], by incorporating Contrastive Learning techniques. We ex- plore the integration of dictionaries, Contrastive Learning, and Large Language Models (LLMs) to improve entity recognition, and we examine the use of LLMs for extracting relationships between entities. The result is a streamlined tool designed to identify and normalize entities in Biomedical abstracts and to effectively extract relationships, thereby enabling advancement of Biomedical Information Retrieval systems

Yazar

Buğrahan Şahin

Bu Yayına Nasıl Atıf Yapılır

Buğrahan Şahin (Master Thesis). Using contrastive learning and large language models on biomedical information extraction, 2024, Boğaziçi University.

Anahtar Kelimeler

Lisans

Tüm Hakları Saklıdır

Bu eser belirtilen lisans koşulları altında paylaşılmaktadır.

Boğaziçi University tezlerinden daha fazlası