Awesome Library

[hamelsmu]

Awesome Library

[anttttti]

Nice, similar to what WordSeq produces. The histogram-based length selection could be useful to have. Feel free to contribute :)

Note that wb is still work in progress, and a few bugs have surfaced now with the Mercari Kaggle competition. Some missing bits as well, such as pickling extractors written in cython. The API should be revised so it's as compatible with sklearn as possible. And the documentation improved.

There's also been a lot of work on the DL libraries for preprocessing since I made the first version available. Keras and others have dataset and dataloader packages. WB only has the wordbatch pipeline for text data, and nothing yet for data munging from different formats (.csv, word), or output of features to formats such as the LibSVM feature files.

[hamelsmu]

No need for the caveats. The first step in open source is putting it out there. Don’t be afraid or feel you need to explain! (I work at Github)