Large Language Model Powered Chatbot for Comprehensive Citizens Information Services in Ireland
Year: 2023
Type: dissertation
Abstract: This study explores the development and evaluation of a large language model (LLM)powered chatbot that employs GPT-4 to improve public information accessibility through WhatsApp. The chatbot uses the Citizens Information Ireland website as a data source. Creating the evaluation dataset is an automated process that ensures that the sample is unbiased and diverse. The chatbot was evaluated using ROUGE, BLEU, BERTScore, and UNIEVAL metrics based on 114 distinct question-answer pairs. A comprehensive analysis of the chatbot's performance was obtained based on mean scores and standard deviations. It demonstrates its ability to provide relevant, accurate, and context-appropriate responses. With a particular focus on improving public information accessibility, this study provides valuable contributions to the literature on LLM-enabled customer support, demonstrating the broader commercial potential of LLM-powered chatbots.
Source: NORMA
Author: Zanello Klostermann, Reinaldo
Language: English
Topic: AI in Service Interactions
Subfield: Marketing
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7039148661