How we built a verified dataset of 198 tour group headset rules

Museums and historic city centers have been implementing various rules regarding tour guide interactions with groups. Some rules mandate headset usage for groups of a certain size, while others prohibit the use of loudspeakers or allow only the site's own guides to narrate indoors. Initially, a survey of 228 candidate rules was conducted, but this list was deemed unreliable due to sources such as travel blogs and outdated rule versions.

To address this, the process was refined to verify each rule by cross-referencing the official source, which involved using LLM agents to search official websites, PDFs, and archived copies of pages. The verified rules were then compiled into a structured dataset, with important fields including the rule class, threshold group size, device policy, and the source URL.

After rigorous filtering based on criteria such as verdict confirmation, source URL validity, and quote presence, 198 rules from 36 countries were finalized. The dataset, categorized into three classes (A, B, and C), is open under CC BY 4.0 and can be accessed on platforms like Hugging Face and Kaggle. The dataset highlights the varying interpretations of headset rules, with class A requiring headset usage for all guided groups, class B prohibiting loudspeakers while allowing headset usage for large groups, and class C restricting indoor narration to site guides or audio guides.

The dataset also includes fields for threshold group size, guide inclusion in the count, device policy, and the source URL for verification purposes. The creation of this dataset aimed to provide a comprehensive, verified reference for tour guides, with plans to periodically update the dataset as rules evolve.