Janis Lesinskis

As a software engineer who's very interested in engineering mathematics I'd be very keen to have a chat with everyone about this project and the directions it is going in....

Is the main blocker to updating to Python 3 maintaining Python 2.7 support for existing users?

The spirit of these things is really about merging in changes from other people but I don't mind running you through some of the topics if you want to do...

I wouldn't say your repo is entirely against the spirit of things, you shared your code here which is highly useful to a lot of people (myself included) so thanks...

This looks like a bit of a bigger task than I first thought because it implies that a fix will involve figuring out how to deal with `speakers` then implementing...

Appreciate the feedback here. "By default, the acoustic model created is speaker-specific, right?" while the library can do this I don't know that the implementation as it stands does this...

@oadams: > For instance, as part of efforts to improve the Na corpus, I recently added a tag to passages said by the investigator, so that they can be automatically...

I propose the `Corpus` keep a record of who the speakers are but that the base `Corpus` doesn't do the processing on the speakers. If this is parameter is not...

Given the ability to infer the speakers from the utterances I think we should get rid of the speakers parameter in the base Corpus.

That call to construct the class is particularly problematic, `label_segmenter` simply cannot be `None` here and hence the default parameter seems broken.