Below we try to estimate the impact of Dr. Povey’s incident on the development of Kaldi, partially based on communications with Dr. Povey.

Kaldi as an Open-Source Software

Kaldi is an open-source software, and we expect most of Kaldi’s current development will remain intact. Dr. Povey has expressed on various occasions that he would only consider employment opportunities that would leave him with enough time to continue to support Kaldi and work with his students. So at least in the near future, we expect Kaldi will keep its current development momentum.

Over the past several years, Kaldi was largely maintained by Dr. Povey, with the help from his students at the Center for Language and Speech Processing, the Johns Hopkins University. Dr. Povey’s students have been a very stable, very effective group of Kaldi developers over the years, and have made tremendous contribution to Kaldi. With Dr. Povey leaving the Johns Hopkins University, his current students will eventually graduate, and there will be no more future students. This may hurt Kaldi’s development in the long run.

Kaldi and the Johns Hopkins University

Kaldi and the Johns Hopkins University (or more precisely, the Center for Language and Speech Processing, CLSP) go back a long way since its inception. Kaldi began its existence in the 2009 CLSP Summer Workshop titled “Low Development Cost, High Quality Speech Recognition for New Languages and Domains”, and has been largely maintained by CLSP faculty and students since Dr. Povey joined the Johns Hopkins University in early 2012. A lot of Kaldi’s lastest features were developed at the Johns Hopkins University, for example, Kaldi’s neural network 3 framework, time delay neural network models, chain models, etc.

With Dr. Povey leaving the Johns Hopkins University, researchers from the Johns Hopkins University can continue to use and develop Kaldi, but it is likely that the Johns Hopkins University will no longer be Kaldi’s development powerhouse.