A Benchmark Comparison of a Domain-Focused Pipeline with ChatGPT
Document Type
Article
Keywords
artificial intelligence; natural language processing; question generation; question answering; financial technology; knowledge discovery; text-based models
Publisher
Journal of Computer Information Systems
Rights Management
Copyright © 2025 Informa UK Limited
Abstract
This paper presents a framework for generating domain-specific questions and retrieving answers from a target-domain text corpus. The approach involves fine-tuning a model to deliver concise answers tailored to mutual funds in finance. Using open-source tools and datasets, the proposed pipeline achieves accuracy levels comparable to ChatGPT, while offering key advantages such as customizable domain-specific corpora, reduced training time, and lower costs. Benchmarking results highlight its effectiveness and potential as a cost-efficient alternative for domain-focused question answering.
