A role-based NLP sentiment analysis of acceptance, trust, and action in autonomous mobility

The Published Research Data for “Who moves with the robotaxi? A role-based NLP sentiment analysis of acceptance, trust, and action in autonomous mobility” provides the full evidence base and intermediate analytical outputs underpinning the role-based findings reported in the manuscript (including the stakeholder-specific action drivers and barriers summarised in Table 2). The repository is organised into three role-stratified folders, Pedestrian_Filter, Passenger_filter, and Motorist_Filter, corresponding to discourse produced by pedestrians, passengers, and motorists, respectively. Within each role folder, we deposit (i) the raw public comment corpus used in the study (with direct identifiers removed and records retained in a structured, inspectable format), (ii) pre-processed text files used for modelling (e.g., cleaned/normalised strings and analysis-ready fields), and (iii) the node-level outputs generated by the NLP pipeline that enable a transparent audit trail from data to interpretation (e.g., per-comment role assignment, sentiment/stance signals, semantic representations, and cluster membership/weights). We additionally include the role-specific thematic cluster tables and the corresponding representative expressions/exemplar snippets used to illustrate each theme, allowing readers to evaluate how frequently a theme appears, how it is linguistically expressed, and how it varies by stakeholder role. A top-level README and data dictionary describe file structure, variable definitions, and decision rules, enabling independent verification of the presence and prevalence of the themes discussed and reducing concerns about selective exemplification. In short, the dataset supports traceability by linking each interpretive claim in the manuscript to the underlying role-specific clusters and illustrative expressions, while providing sufficient intermediate outputs for readers to reproduce and scrutinise the core analytical steps.