TRAINS Dialogue Corpus

Home | People | Projects | Publications | Resources

As part of the TRAINS project, a corpus of problem solving dialogues was collected. The dialogues involve two participants: one who plays the role of a user and has a certain task to accomplish, and another who plays the role of the system by acting as a planning assistant.

The first collection

The first dialogue collection was done for the purpose of studying the prosody of discourse structure, and is briefly documented in the technical note ``A study on prosody and discourse structure in cooperative dialogues'' ( text , figures ). The collection was not very extensive, but it was the first one using the Trains domain.

The Trains 91 Dialogues

The second set of dialogues were collected in 1991. The manner in which these dialogues were collected is described in the technical note ``The Trains 91 Dialogues''. The transcriptions of these dialogues are also available.

The Trains 93 Dialogues

In 1992 and 1993, more dialogues were collected, this time using the Waves software (Entropic Research Laboratories), which allows time-aligned word transcriptions to be produced. These dialogues along with the word transcriptions, are available on CD-ROM from the LDC. Altogether, this corpus includes 98 dialogs included, collected using 20 different tasks and 34 different speakers. This amounts to six and a half hours of speech, about 5900 speaker turns, and 55,000 transcribed words. The collection and transcription of the dialogues is documented in the technical note ``The Trains 93 Dialogues''. The transcriptions themselves are available here.

Dialogue Transcription Tools

In collecting and annotating ``The Trains 93 Dialogues'', we developed a set of tools for converting a DAT recording into a fully segmented and annotated dialogue. These tools allow the user to progress stepwise through this process, from creating the initial dialogue audio file, breaking up the dialogue into a sequence of single-speaker utterance files that preserve the sequentially of the dialogue, annotating the utterance files, printing the contents of the dialogue, and updating the breakup of the dialogue. These tools are described in the Trains technical note, ``Dialogue Transcription Tools'', and will be available through ftp, as well as on the CD-ROM. The toolset itself is available in a tar file.

Uses of the Corpus

The dialogues that we have been collecting have played an integral part in the Trains project. They have supplied us with empirical evidence of the types of phenomena that can occur in spoken dialogues. They have also provided us with data for our empirically based work. For instance, we have used it to train a parser that uses statistical preferences, and to train a part-of-speech tagger that models speech repairs (Heeman and Allen, 1994).

Last change: 1 Dec 2000 by stent