IARPA said Thursday the ASpIRE challenge launched in October 2014 sought speech recognition platforms that can transcribe a recorded speech and work on single microphone and multiple microphone recordings.
A group of individuals from the Institute for Infocomm Research at Singapore-based A*STAR won in both single- and multiple-microphone categories.
IARPA also selected a team from the Center for Language and Speech Processing at Johns Hopkins University and another group composed of researchers from Johns Hopkins University, Raytheon BBN Technologies and Czech Republic-based Brno University of Technology for the single microphone category.
“We’re delighted with the diversity of solutions submitted by the ASpIRE challenge contestants,” said Mary Harper, program manager for the ASpIRE challenge at IARPA.
“Their performance under rigorous evaluation conditions suggests that accurate speech recognition – even for speech recorded in environments for which training data are unavailable – is possible.”
IARPA and the Massachusetts Institute of Technology’s Lincoln Laboratory assessed the results.
The agency also collaborated with the Linguistic Data Consortium for speech data collection, Appen Butler Hill for the transcription of microphone recordings and InnoCentive for the management of ASpIRE challenge website.