Janis Lesinskis

https://github.com/CoEDL/elpis/blob/913c567bfaafc2081bc96fbd80036b07d4762bf0/elpis/endpoints/__init__.py#L26-L33 From the point of view of making tooling that interacts with this project it would be easier to deal with a missing log file if the following code had...

https://github.com/CoEDL/elpis/blob/913c567bfaafc2081bc96fbd80036b07d4762bf0/elpis/endpoints/__init__.py#L28-L30 There's some subtle bugs that can occur here if the file encoding is not specified. I'd suggest making this explicit.

There is currently some code that returns an empty list and has a TODO, it is likely best to raise an exception here so that incomplete code cannot slip past...

Improve the default for the Kaldi path such that it does not reference an explicit users home directory. Update documentation. Closes #237 Closes #239

Currently there's a mismatch between how the settings.ini file is loading the Kaldi root path and how the documentation explains the same. The documentation expects to read in the key...

It might be good to provide some links to papers that have used Persephone on the https://persephone.readthedocs.io/en/stable/ site

[WIP] Make Keras model

1

Create a model for RNN CTC using Keras. Note: Use of Keras save functionality will greatly help in saving and restoring models.

Test for Elan corpus construction

3

Would be good to have a small integration test for constructing a corpus from an Elan file for coverage reasons, see #177

Corpus.from_elan fails with default argument for label_segmenter

8

`Corpus.from_elan` has a parameter that will allow you to specify a label segmenter which is of type `LabelSegmenter` Usages: ```python utterances = [label_segmenter.segment_labels(utter) for utter in utterances] ``` and ```python...

Segmentation of characters with unicode space characters

2

Right now Unicode space characters such as 'NO-BREAK SPACE' (U+00A0) don't get split on. I propose we decide on what behavior we expect here and resolve this in PR #213....