If ZeroSpeech will be held in 2021 as every year, it's good to practice with the setting which was used for 2020 to get the picture. So, this repository is made in order to (i) train NPB-HLM based on the data-set (ii) compare the model with the baseline (iii) evaluate and submit the results for the purpose of practice.
For more general background information, see http://zerospeech.com/2020/ .
submission: http://zerospeech.com/2020/instructions.html#registration