Supervised Semantic Similarity-based Conflict Detection Algorithm: S3CDA

Abstract

Identifying conflicting requirements is a key challenge in software requirement engineering, often overlooked in automated solutions. Most existing approaches rely on handcrafted rules or struggle to generalize across different domains. In this paper, we introduce S3CDA, a two-phase algorithm designed to automatically detect conflicts in software requirements. Our method first identifies potentially conflicting requirement pairs using semantic similarity, and then validates them by analyzing overlapping domain-specific entities. We evaluate S3CDA on five diverse real-world datasets and compare it against popular large language models like GPT-4o, Llama-3, Sonnet-3.5 and Gemini-1.5. While LLMs show promise, especially on general datasets, S3CDA consistently performs better in domain-specific settings with higher performance. Our findings suggest that combining Natural Language Processing (NLP) techniques with domain-aware insights offers a practical and effective alternative for conflict detection in requirements.