Sidebar Menu

Team

   Professors
   External Lecturers
   Office Assistance
   System Administration
   Internal PhD Students
   External PhD Students
   Students
   Alumni
   Company Founders
   DSS Cups

Research

   Research Topics
   Real-time Framework
   Publications

Teaching

   Lectures
   Labs
   Seminars
   Student Projects
   Theses
   Evaluations
   Exams
   RED

News

Events

   Talks

Media Center

   Audio Examples   
   Data Bases
   Surveys

GaS

   Basic Information
   Prices
   Members
   Reports
   Statute

Map Task

 
  Contents:

The Map Task scenario was used to elicit spontaneous speech for the SPID corpus. Both dialogue partners received a map, with the instruction giver having a route drawn on it. The other speaker was the instruction follower. The giver had to explain the route to the follower, who had to draw it on his/her own map. Different maps were designed for each of the recording conditions. To allow further analyses of the influences of different maps on the speech signals, 6 maps were only used in the laboratory conditions, and 4 of them only for recordings in the real driving situations. Seats and roles of giver and follower were chosen by the subjects, but never changed within each dialogue pair for the rest of the experiment.

The two maps given to a pair of subjects shared many common features, but there were some keywords that only appeared on the giver’s map or on the follower’s map. In addition, names of streets and persons as well as house numbers were slightly different. Some parts of the maps even had slightly different structures. Very often, the giver had a name on his/her map like “Schulz”, whereas the follower had two names, “Schulz” and “Scholz”, so that a controversial discussion was fostered. Maps and noise conditions were randomised for both recordings in the laboratory and the driving car.

 
Map of the follower.   Map of the giver.