Matthias

Hi, if you like you can implement a reader for another dataset/corpus from the list above or any other one you know. You can find examples of existing implementations in...

Yes, that is a good idea. I also had something similar in mind. Due to some datasets changing "frequently", I wanted to introduce versions. So you could actually select which...

Did your problem occur, when executing the scripts as they are? What exactly have you changed to get relative paths like ``../../megs/data/download``?

Did you also run the waverize script before exporting to kaldi?

Ok, so the consistency check is just for making sure having the same data. This is not that important for the rest to work at all. And it is also...

The main problem is, that all utterance durations are needed in the kaldi files. It needs to get these from the audio files. Therefore it takes that long.

With utt2dur you are right, but it is also needed for the segments file, since some of the utterances only are segments of the full audio file. Have you tried...

Yes, good point. I think public AppVeyor would be a good choice.

Or is there any difference to the private AppVeyor? Can "outsiders" see the build reports?

Would definitely make sense to run benchmarks as well. We could try it on our AppVeyor instance, as soon as we find time to set it up.