I am looking for a dataset of clear audio, a corresponding transcript (optional), and most importantly a list of all the phonemes said in the audio, with the length of each phoneme and a mention of each phoneme start and end time. I looked through stuff like gentle forced aligner, however, the problem is that there is a small offset that will be very bad to use for my project.

To clarify, my project is a convolutional neural network that can classify a small spectrogram into a phoneme.

  • $\begingroup$ The kind of data is like gold dust, so I don't think anything like it will be freely available. It is very time-consuming to annotate such data. A principal problem is also that you want to identify discrete units in a continuous stream, so star/end time and length doesn't really apply: the sounds merge into each other when speaking. $\endgroup$ Dec 6 '19 at 9:06
  • $\begingroup$ So my project is trying to transcribe all possible words (in a certain dictionary) spoken by anyone. How do I approach this problem? $\endgroup$ Dec 6 '19 at 13:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.