Loqui Human-Human Dialogue Corpus (Transcriptions and Annotations)

The testbed application, the CheckItOut dialogue system, was modeled on a corpus of telephone transactions between patrons and librarians that we collected at New York City’s Andrew Heiskell Braille & Talking Book Library in 2006. This data release contains transcriptions of the original telephone transactions of eighty two dialogues that were collected and used to inform the initial design of CheckItOut.

It also contains annotations that capture dialogue acts, adjacency pairs (e.g., links between questions and their answers), discourse units, and specificity of referring expressions about the books under discussion.