On this webpage we share and make publicly available a corpus of transcribed human-human dialogs in German language provided in PDF-format.
The corpus was created in order to derive linguistic heuristics from human-human communication and adopt it to human-robot dialog in the context of asking for directions. It is intended for research use and the corpus is a free documentation released under the terms of the GNU Free Documentation License (GFDL). You may copy or reproduce and use the corpus for your own research. Scientific publications based on the corpus are requested to acknowledge the source:
Barbara Gonsior and Andrea Bauer, Wegauskünfte in München, Technische Universität München, 2010 www.lsr.ei.tum.de/fileadmin/publications/MuenchenKorpus_2009.pdf
The 48 asking-for-directions-dialogs have been recorded in summer 2009 in the city of Munich asking passers-by for the way to three different goals on five different routes. In order not to influence the participants and gain natural dialogs the participants were not informed about the recording of the conversations. Exceptions are dialogs 4, 6, 20 and 34, in which the subjects detected the recorder while describing the route. Afterwards, all participants agreed that the dialog-transcription could be published in anonymous form. Accordingly, the dialog partners are marked “F” for the asker and “A” for the responder, more than one responding passer-by are marked “A1, A2 …”. Overlapping utterances are not identified within these transcriptions. The starting point and the designated goal location are indicated above each dialog. Additionally, the asker’s initial walking orientation is quoted, as well as the direction of the mutual co-orientation of both conversation partners in the course of the dialog.