Babel program

{{Short description|Speech recognition technology for noisy telephone conversations}}

The IARPA Babel program developed speech recognition technology for noisy telephone conversations. The main goal of the program was to improve the performance of keyword search on languages with very little transcribed data, i.e. low-resource languages. Data from 26 languages was collected with certain languages being held-out as "surprise" languages to test the ability of the teams to rapidly build a system for a new language.{{cite web|last1=Harper|first1=Mary|title=Data Resources to Support the Babel Program Intelligence Advanced Research Projects Activity|url=https://www.ldc.upenn.edu/sites/www.ldc.upenn.edu/files/harper.pdf|accessdate=26 July 2017}}

Beginning in 2012, two industry-led teams (IBM and BBN) and two university-led teams (ICSI led by Nelson Morgan and CMU) participated.{{cite web|title=Babel|url=https://www.iarpa.gov/index.php/research-programs/babel|website=IARPA|accessdate=26 July 2017}} The IBM team included University of Cambridge and RWTH Aachen University, while BBN's team included Brno University of Technology, Johns Hopkins University, MIT and LIMSI. Only BBNT. Alumäe et al., "The 2016 BBN Georgian telephone speech keyword spotting system," 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, 2017, pp. 5755-5759, doi: 10.1109/ICASSP.2017.7953259.

and IBMJ. Cui et al., "Knowledge distillation across ensembles of multilingual models for low-resource languages," 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, 2017, pp. 4825-4829, doi: 10.1109/ICASSP.2017.7953073.

Gales M.J.F., Knill K.M., Ragni A. (2017) Low-Resource Speech Recognition and Keyword-Spotting. In: Karpov A., Potapova R., Mporas I. (eds) Speech and Computer. SPECOM 2017. Lecture Notes in Computer Science, vol 10458. Springer, Cham. https://doi.org/10.1007/978-3-319-66429-3_1

P. Golik, Z. Tüske, K. Irie, E. Beck, R. Schlüter, and H. Ney. The 2016 RWTH Keyword Search System for Low-Resource Languages. In International Conference Speech and Computer (SPECOM), Lecture Notes in Computer Science, Subseries Lecture Notes in Artificial Intelligence, volume 10458, pages 719-730, Hatfield, UK, September 2017.

made it to the final evaluation campaign in 2016, in which BBN won by achieving the highest keyword search accuracy on the evaluation language.

Some of the funding from Babel was used to further develop the Kaldi toolkit.{{cite web|title=History of the Kaldi project|url=http://kaldi-asr.org/doc/history.html|accessdate=26 July 2017}} The speech data was later made available through the Linguistic Data Consortium at a symbolic cost of $25 USD per language pack.