Abstract

Accessing complex database information often requires specialized knowledge. Existing dash\-board tools supporting only pre-defined queries are inflexible and inadequate for many users. This paper introduces a novel approach by combining an intuitive, speech-based interface with a Large Language Model (LLM) specifically trained to understand database structures and generate user-friendly answers and visualizations. Our method focuses on simplifying the interaction for non-expert users by allowing them to ask questions directly and receive insights without the need for database engineers or data scientists. This is especially important to handle follow-up questions raised by anomalies in generated answers. We present a detailed analysis and discussion of a use case that demonstrates the practicality and effectiveness of our approach through a developed prototype. We study the strengths and weaknesses of the system components as well as the user feedback for the system. Based on the observations further research directions are discussed.

@inproceedings{KiwanEtAl-InteractiveDataVisualisation,
author = {Abdullah Kiwan and Andreas Lommatzsch and Sahin Albayrak},
title = {Enhancing Data Accessibility: Integrating Speech-Based Interfaces with Large Language Models for Intuitive Database Queries},
booktitle = {Proceedings of the Conference Lernen, Wissen, Daten, Analysen},
numpages = {12},
year = {2024},
series = {LWDA '24},
location = {W\"urzburg, Germany},
note = {urn:nbn:de:0074-XXXX-X},
issn = {1613-0073},
publisher = {CEUR Workshop Proceedings},
url = {https://ceur-ws.org/Vol-XXXX/LWDA2024-paperXX.pdf}
}