search

aalto-speech / AaltoASR

Aalto Automatic Speech Recognition tools
Other
85 stars 37 forks source link

rectool.py example use #21

Open gabrielecastellano opened 8 years ago

gabrielecastellano commented 8 years ago

Hi, i would like to use AaltoASR python to perform a recognition starting from an audio file as input, but i can't figure out how to inizialize some variable of the example bash (as AM, LM and so on...). Where can i find a detailed documentation or working example?

Thanks in advance.

psmit commented 8 years ago

The AaltoASR speech recognition system needs an acoustic model (AM) trained with AaltoASR and a ngram language model trained with e.g. srilm. Do you have any experience training ASR systems? Also, do you happen to be an Aalto University student, in which case we can see what models we can provide for you?

gabrielecastellano commented 8 years ago

I'm from Politecnico of Torino (IT), actually working for this research https://github.com/netgroup-polito

psmit commented 8 years ago

Unfortunately our external documentation for AaltoASR is not very good. Unless you are specifically interested in the features of our decoder (token-pass, good for deep-context subword models) I can recommend to use a different speech recognition toolkit such as Kaldi or HTK. If you do very much want to use our decoder, let me know and we'll try to get you started.