HCRC Map Task Corpus

The HCRC Map Task Corpus consists of audio, transcription with hand-timestamped word-level timings, and annotation for 128 5-10 minute task-oriented dialogues. The dialogues are elicited using a map navigation task that results in very natural conversation whilst still exerting enough control to allow meaningful comparisons for a number of variables that were built into the design. The speakers are Scottish students. The annotations, which cover a wide variety of mostly discourse phenomena, are in the format for the NITE XML Toolkit.

• url: http://www.hcrc.ed.ac.uk/maptask/
• main_author: The Human Communication Research Centre, University of Edinburgh
• license: Creative Commons Non Commercial Attribution Share-Alike
• subjects: 64
• recordings: 128
• duration: 5-10 minutes
• naturality: roleplay
• media: close-talking microphone output for each speaker spliced into a stereo file, one speaker per channel; the associated scanned maps. Some video exists but does not form part of the release.
• language: English
• interaction: dyadic
• annotation: manually aligned transcripts