Kadot

PositionalWordVectorizer is the vectorizer I built for named entity recognition (or similar tasks). It can help better understand the context of a word by taking in count the proximity with other words.

Some other corrections included.

This version add a new classifier : Naives Bayes and a simpler document vectorizer : a count based bag-of-word model.

These two **basics** of NLP are now in Kadot !

(Note that this release is a bit small, but, just by his name, it solve a part of this [issue](https://github.com/the-new-sky/Kadot/issues/2))

This update delete `VectorCoordinate` class and use `np.array`. A much complete readme was added and `apply_vector` method was renamed by `apply_translation`.

This is a very cool update. Better vectorizers, ngrams, apply_vectors...

Sorry, I'm a noob... 2H after trying and retrying to push Kadot on PyPi, it seams ok !