4 ms·
Open source Kaldi gives you 7.8%, Microsoft didn't went too far. Also, major issue with this kind of research is that they combined several systems in order to
by nshm 10y ago
Open source Kaldi gives you 7.8%, Microsoft didn't went too far.
Also, major issue with this kind of research is that they combined several systems in order to get best results. Most practical systems don't use combinations, they are too slow.
- danielmorozoff 10y agoCould you please provide a link?
- nshm 10y agohttps://github.com/kaldi-asr/kaldi/blob/master/egs/fisher_swbd/s5/RESULTS https://github.com/kaldi-asr/kaldi/blob/master/egs/fisher_sw... In Microsoft paper http://arxiv.org/pdf/1609.03528v1.pdf http://arxiv.org/pdf/1609.03528v1.pdf Table 5 it's a line "Povey et al. [19] LSTM". http://www.isca-speech.org/archive/Interspeech_2016/pdfs/0595.PDF http://www.isca-speech.org/archive/Interspeech_2016/pdfs/059... Interpolate it to RNNLM column and you'll get 7.8. See also http://www.isca-speech.org/archive/Interspeech_2016/pdfs/0473.PDF http://www.isca-speech.org/archive/Interspeech_2016/pdfs/047...
- danielmorozoff 10y agoCould you please provide a link?
- WilliamDhalgren 10y agoSo this model won't be free software??? Odd, and bummer... Also I'm note sure an error reduction of 20% (1-6.3/7.8) is to be considered small; depends on the particular challenge really. Like, sentiment analysis only starts to get interesting above 80% on some dataset, as much can be guessed correctly in very naive ways.. Human lvl on this task is estimated to be ~4% so we have quite a lot of ground to cover still..
- visarga 10y agoMaybe they can distillate the ensemble in a compact and efficient version for production.
- timgws 10y agoThere are 33% less errors with the Microsoft solution then with Kaldi... one could say that is quite significant.