Group Affect and Performance (GAP) Corpus
The GAP corpus is released under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. https://creativecommons.org/licenses/by-nc/4.0/
Below is a description of the GAP data. All conversations are in English.
Meeting and Group Member Identification (ID) Codes: For all corpus components, meetings and group members are identified with a code. Meetings are coded with a number based on the temporal order of data collection. Group members are coded with one of five arbitrarily assigned colors (i.e., blue, green, pink, orange, yellow).
Meeting Transcripts: Meeting transcripts are contained in the "Transcripts/" folder in .txt files. Each transcript file is titled with the corresponding meeting's ID number and data/time of data collection. Each utterance is labeled with the group member's ID, the utterance number for that group member, and the start and end time stamps of the utterance.
Meeting Annotations: Meeting annotations are contained in the "Annotations/" folder. This includes annotation of decision-making utterances, sentiment-bearing utterances, utterances that refer to a participant's individual ranking, and utterances that refer to a survival item. Each annotation contains the ID, number, and start and end time stamps that correspond to the annotated utterance.
Merged Transcripts and Annotations: Files which merge the transcripts and annotations can be found in the "Merged/" folder.
Meeting Data: Winter survival task data, post-task questionnaire responses, participant characteristics, and meeting characterists are contained in the "Group-Individual-Data/" folder. One Microsoft Excel file contains the group-level data i.e., "Group-Level Meeting Data". Meeting size, meeting length, Absolute Group Score on the winter survival task (AGS), Group Time Expectation (Group_TE), Group Worked Well Together (Group_WW), Group Time Management (Group_TM), Group Efficiency (Group_Eff), and Group Quality of Work (Group_QW), and Group Overall Satisfaction with Meeting (Group_Sat) are provided. One Microsoft Excel contains the individual-level data i.e., "Individual-Level Meeting Data". Year at the university (Year at UFV), Gender, whether English is their first language (English; 1 = yes; 2 = no), Absolute Individual Score on the winter survival task (AIS), Absolute Individual Influence on the winter survival task (AII), Individual Time Expectation (Ind_TE), Individual Worked Well Together (Ind_WW), Individual Time Management (Ind_TM), Individual Efficiency (Ind_Eff), and Individual Quality of Work (Ind_QW), Individual Overall Satisfaction with Meeting (Ind_Sat), and Individual Leadership (Ind_Lead) are provided. Group members are coded with their ID code and meeting ID code (e.g., 1.Blue for participant Blue in meeting 1; 3.Pink for participant Pink in meeting 3).
Meeting Audio: The audio recordings (.wav files) are available as a separate download from the corpus webpage. Each meeting recording is titled as the meeting's ID number and date/time of data collection.
For an in-depth discussion of transcription, annotation, the scores, and how they were derived, please refer to corpus release paper PDF on the corpus release webpage.