A systematic mapping study of information retrieval-based requirements traceability methods

Requirements traceability (RT) is critical for ensuring consistency, quality, and maintainability in software development. While learning-based approaches have gained increasing attention, traditional information retrieval (IR) methods remain widely used in practice. However, existing literature lacks a systematic synthesis of their best practices and recent advancements. To address this gap, we conducted a systematic mapping study (SMS) of 40 primary studies published between 2014 and 2024, selected from an initial pool of 2,052 publications. Our review examines widely adopted IR models, enhancement strategies, evaluation datasets, performance metrics, and baseline methods. Specifically, we identify and categorize 32 representative enhancement strategies into four methodological types: (1) artifact text information, (2) artifact structural information, (3) model-based optimization, and (4) human intervention. Furthermore, we analyze 53 commonly used datasets and 9 evaluation metrics for validation. Our findings indicate that among various IR models, the Vector Space Model (VSM) and Latent Semantic Indexing (LSI) typically achieve stronger performance in RT tasks. This study provides a comprehensive synthesis of IR-based RT research and offers practical insights to advance traceability in software engineering.