CoFee data set is grounded on four corpora. All these corpora have been transcribed and their transcriptions have been force-aligned at phone level:

  • CID : Corpus of Interactional Data (A conversational corpus created by Roxane Bertrand and Béatrice-Priego Valverde), 16 speakers, 8 hours
  • MTR : Maptask-Aix (a French Maptask created by Corine Astésano and Ellen Gurman Bard), 8 speakers, about 2h30.
  • MTX : Audiovisual maptask (A replication of the MTR corpus in a face to face condition), 10 speakers, about 2h30. This corpus had been recorded and created within CoFee project.
  • DVD :  AIX – DVD (A discussion and argumentation corpus about movies, 8 speakers, about 4 hours. This corpus had been recorded and created within CoFee project.

The primary data of these corpora as well as the annotation produced in the project are accessible under an SLDR CoFee collection.

Automatic annotation produced on the raw data that are present in the data set are:

  • Various acoustic features extracted

Manual annotation produced on the data set are:

  • Cross-coded annotation of potential feedback utterances composed of specific lexical items and pre-segmented facial expressions and head movements (N=9428)



Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.