Dataset

This dataset contains dialogues between a dialogue system and Amazon Mechanical Turk Users collected between 2011 and 2018. Each dialogue contains the output of automatic speech recogniser, respective system input and the user feedback.

Every directory is one experiment. In each experiment directory there are a number of directories with names of the following format: voip-USERID-STARTTIME_ENDTIME Each of these directories represent one dialogue. In each dialogue directory session.xml - dialogue log session.cfg - system configuration file feedback.xml - user feedback database.txt - ontology file rules.txt - ontology file In session.xml <systurn> - system turn <userturn> - user turn <asrhyp> - asr hypothesis with confidence score <dact> - system dialogue act <semihyp> - semantic user hypothesis In feedback.xml goal: describes the user goal task: goal in natural language question: contains the feedback database.txt - contains dialogue entities that the dialogue system can talk about rules.txt - describes all concepts that the dialogue system can talk about