Natural Language Dialogue group

Transcripts of a large scale training exercise from Fort Rucker, involving multiple teams of trainees using flight simulators, exercising a coordinated mission with a command post and semi-automated simulated forces. The exercise runs 80 minutes, but contains simultaneous speech over multiple radio channels, as well as side communication between team members in the same location, captured over an open channel, yielding nearly 10 hours of multimodal speech by 38 speakers in total.

The corpus is fully coded at the utterance level with sentence structure, dialogue acts, grounding acts, communication management, intonation, addressee and modality (radio or local). In addition, 'dialogue episodes' are defined and annotated with one of 18 activity labels based on the dialogue's function and general orientation to the larger task.