The speech recognition virtual kitchen turns one

This paper describes recent developments in the Speech Recognition Virtual Kitchen. The core of the research infrastructure is the use of Virtual Machines or Linux containers. We liken these to a “kitchen”, because they provide “appliances” (e.g., speech recognition tool-kits), “recipes” (scripts for creating state-ofthe art systems), and “ingredients” (language data). Users can now download a number of pre-configured Virtual Machines at http://speechkitchen.org/, and open-source versions will be made available on Github soon. Several workshops and evaluation campaigns are distributing baseline systems as Virtual Machines, furthering the goals of promoting community sharing of research techniques, fostering innovative experimentation, and providing solid reference systems as a tool for education, research, and evaluation in the speech and language area. A web-based community platform complements the VMs, to allow physically disconnected users to jointly explore VMs, learn from each other, and collaborate in research