From Speech to Safe Action: Research Dataset and Evidence Matrices

This dataset accompanies the manuscript entitled "From Speech to Safe Action: A Structured Review of Voice Interaction with Industrial Mobile Robots and in Adjacent HRI Domains". The study presents a structured literature review of AI-supported voice interaction in industrial mobile robotics and related human–robot interaction (HRI) domains, focusing on task performance, human factors, acoustic robustness, and physical safety. The curated corpus comprises 122 publication records. The main synthesis includes 74 publications (7 Direct and 67 Transferable), representing 66 publication families and 170 evaluation units. The remaining corpus contains 15 Context publications, 27 reviews and design-guidance publications, and 6 excluded records. The accompanying files provide publication-level classifications, evidence matrices, safety outcome assessments, comparisons with previous reviews, independent coding verification results, and documentation of the targeted literature update. The dataset supports transparency, verification of reported findings, and further research into the effectiveness, ergonomics, robustness, and safety of intelligent voice interfaces for robotic systems. The review employs a structured, descriptive synthesis rather than a fully reproducible systematic search or meta-analysis. Publication counts should not be interpreted as estimates of prevalence across the entire research literature. The dataset does not redistribute full-text publications from third-party publishers. Detailed methodological definitions, eligibility criteria, analytical procedures, and interpretation limitations are documented in the accompanying manuscript and supplementary material.