Utterance separated into phonetic level?

[Lose Looser]

I am currently using the an4 dataset available from here
I am trying to do phone recognition, but the utterance has only the words and not the phones. Is there someway i can extract utterance which is phonetically seperated?

[Nickolay V. Shmyrev]

Acoustic model is the same for phone and words recognition. You do not need a database separated on phonemes for phonetic recogniton, you can use conventional database. If you need higher accuracy you can use larger dataset like tedlium for training.

[Lose Looser]

But it only seems possible to do it on the test set.. rather than the train set..

[Nickolay V. Shmyrev]

You need to be more clear what do you mean by "it".

[Lose Looser]

I tried converting my utterances using the acoustic model to phonetic level.
What i ended up getting was only the test set seperated into phonemes.

[Nickolay V. Shmyrev]

And what is the problem? Until you explain in details nobody will help you.

[Lose Looser]

There is two problems.. I am not sure why but I was only able to do it using the test set..
Second problem being that the number of phone classes. Most phoneme recognition describe the issue as a 61 class problem, but i seem to have more than 61 classes. above 100?..

[Nickolay V. Shmyrev]

In Kaldi group you provided much better description, it is sad you are trying to fool us here. No much to discuss then.