Beam decoder only supports single letter labels?

[weedwind]

Hi,

I am using CTC to do phoneme recognition. My labels are phones, such as ["AX", "OW"] etc. I tried to use the basic beam decoder (without lm) to do decoding. But the decoder reports that the label size did not equal the probability tensor size. If I change to label to single letters, like ["A", "B"], it was fine. So I am wondering how to deal with multiple letter labels?

Thank you very much

[SilvrDuck]

This is apparently because of the implementation of the underlying ctcdecode library. It doesn’t look like it will be supported for now.

See here parlance/ctcdecode#31

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.