Output details
15 - General Engineering
University of Cambridge
Morphological decomposition in Arabic ASR systems
This work was funded by the US DARPA-funded program, Global Autonomous Language Exploitation (GALE). This paper details the incorporation of a standard Arabic morphological decomposition tool "Morphological Analysis and Disambiguation for Arabic" (MADA) into a speech recognition system. The approach formed a central part of the Cambridge University Arabic speech recognition system for the annual NIST-run evaluations. Cambridge was in the BBN team, which achieved the specified targets in the majority of conditions in competition with teams from IBM and SRI. This success led the BBN-led AGILE team receiving funding for all phases of the GALE program.