Multi-Party Open-Ended Conversation Data from a Social Robot System

A supplementary file from a study evaluating a multi-party conversational system for social robots. The system, implemented on a Furhat robot, combines multimodal perception with a large language model and was tested with 30 participants across two interaction scenarios. The PDF document reports results including addressee accuracy and face recognition reliability from experiments conducted by author Giulio Antonio Abbo.

Use Cases

• Benchmarking speaker diarisation and addressee recognition systems based on the reported accuracy metrics.
• Training or fine-tuning LLMs for multi-party, open-ended dialogue based on the conversational data described.
• Analyzing the impact of multimodal cue integration (voice, face) on social presence and engagement in HRI.
• Studying technical challenges like response latency and audio recognition errors in group robot interactions.

Strengths

• Includes specific performance metrics: 92.6% addressee accuracy in parallel settings.
• Reports concrete reliability ranges: 80–94% for face recognition.
• Evaluation based on a defined participant pool: 30 participants across two scenarios.

Limitations

• The primary data file is a PDF (165.2 KB), likely containing analysis and results rather than raw interaction data.
• Row count and column-level documentation for any underlying datasets are unknown.
• Data may reflect the specific experimental setup and technical limitations of the Furhat robot platform.

Provenance

Source

figshare, authored by Giulio Antonio Abbo.

Collection Method

Data likely gathered from controlled human-robot interaction experiments.

Time Range

null

Freshness

Last updated 2026-04-15 04:36:45.

Geography

null