TAXI was produced by BAS, in collaboration with the German research centre for artificial intelligence, DFKI. This speech database contains recordings which consist of dialogues, 94 on the whole (spontaneous speech), between a German speaking cab dispatcher and his clients, who always answered in English. To prevent overlap and to allow automatic segmentation by the recording server, each party pressed a button on his phone to signal the other one that his turn was over. They were recorded over the telephone network. Each dialogue part is translated into the other language. Noise markers are included in the transcripts (not in the translations).The database was annotated following the SpeechDat specifications, and validated to assess its compliance with the SpeechDat format. The files are stored as BAS Partitur Format files.