japaneselat.blogg.se - Spanish speech to text software

‍ Out-of-the-box accuracy is not everything As always, we invite you to review our apps, sign-up and test our accuracy with your data. Our improved releases are even more frequent than that.Īs you can see the field is very close and you get different results on different files (the average and median do not paint the whole picture). Microsoft, on the other hand, releases an improved recognizer every 6 months. Google seems to have the longest development cycles with very little improvement since Sept. You can clearly see that Voicegain and Amazon started quite bit behind Google and Microsoft but have since caught up. (Note for Google the latest result is from latest-long model, other Google results are from video enhanced.) We now have done the same benchmark 4 times so we can draw charts showing how each of the recognizers has improved over the last 1 year and 9 months. Note, the numbers do not add to 63 because there were a few files where two recognizers had identical results (to 2 digits behind comma).

Google Video Enhanced wins a participation trophy by being best on 1 file, which was a very easy "The Art of War by Sun Tzu Full" Librivox Audiobook - WER of 1.79%.

Voicegain was close behind Amazon by being best on 12 audio files.

Amazon was best on 15 files (note that in the October 2021 benchmark Amazon was best on 29 files).

Microsoft was best on 35 out of the 63 files.

Let's look at the number of files on which each recognizer was the best one. Google latest-long, Voicegain, and Amazon are now very close together, while Microsoft is better by about 1 %. The chart also reports the average and median Word Error Rate (WER)Īll of the recognizers have improved (Google Video Enhanced model stayed much the same but Google now has a new recognizer that is better). You can see boxplots with the results above. It was a bad quality phone interview ( Byron Smith Interview 111416 - YouTube). This time only one file was that difficult.

We have repeated the test using similar methodology as before: used 44 files from the Jason Kincaid data set and 20 files published by rev.ai and removed all files where none of the recognizers could achieve a Word Error Rate (WER) lower than 25%. We have decided to no longer report on Google Standard and IBM Watson accuracy, which were always far behind in accuracy. Accuracy of Video Enhanced stayed pretty much unchanged. Google has released a new model "latest-long" which is quite a bit better than the previous Google's best Video Enhanced model.Microsoft and Amazon both improved, with Microsoft improving a lot on the more difficult files from the benchmark set.This resulted in a further increase in the accuracy of our model.Īs far as the other recognizers are concerned: Since then we have obtained more training data and added additional features to our training process. Back then the results were as follows (from most accurate to least): Microsoft and Amazon (close 2nd), then Voicegain and Google Enhanced, and then, far behind, IBM Watson and Google Standard.

It has been over 7 months since we published our last speech recognition accuracy benchmark.