Towards Contextually Aware Large Language Models for Software Requirements Engineering: A Retrieval Augmented Generation Framework
Year: 2024
Type: dissertation
Abstract: This thesis explores the application of Large Language Models (LLMs), particularly open-source ones, to automate and enhance the Software Requirement Engineering (SRE) process. It proposes a Retrieval-Augmented Generation (RAG) framework that empowers LLMs to dynamically access and utilize a knowledge base of relevant documents, enabling them to provide contextually rich and informative responses to Requirement Engineering-related queries. The framework is evaluated through various benchmarking schemes to provide insights into its practical use. It aims to integrate into the SRE workflow, supporting requirement engineers in Q&A tasks related to Software Requirements Specifications (SRS) documents such as requirement clarification, information retrieval, and tracing. The research investigates the effectiveness of the proposed framework in enhancing LLMs for SRE tasks and evaluates the impact of different design choices on its performance. It adopts an experimental approach, employing quantitative methods to analyze the results and compare them against a company-specific ground truth dataset. The motivation behind this research is to improve the efficiency and effectiveness of SRE practices during SRS document-related tasks, ultimately contributing to the development of higher-quality software products.
Source: Doria (University of Helsinki)
Author: Ibtasham, Md Saleh
Institution: Mälardalen University
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7053127786
