Entity linking over tabular data using Large Language Models

Abstract

Entity linking known as the process of linking textual entities with their corresponding entries in knowledge bases. This helps to improve the usability of tabular data. We aim to address the shortcomings of traditional approaches by leveraging advanced language models like openAI or Open source in entity linking task. This will maximize the utility of tabular data across various applications or areas such as data analysis, information retrieval.

We proposed to use OpenAI model GPT-4-0125- preview model for entity linking task. By giving a prompt containing an entry from the table entity, along with its surrounding context and a set of candidate entities, we aim to leverage the contextual understanding capabilities of LLM to accurately associate the entry with a candidate entity from the provided set. We want to analyze the use of LLM for entity linking in tabular data.