Mlconjug3

------------------

* Fixed bug that would silently skip uppercase characters in the CountVectorizer. See: https://github.com/scikit-learn/scikit-learn/pull/19401
* Retrained all models with fixed CountVectorizer.
* Added full support for pyproject.toml file for building the package from source.
* Migrated build system to poetry.
* Updated dependencies.
* Updated documentation.

-------------------

* Fixed typo in requirements specification.

-------------------

* Retrained all models with scikit-learn 1.0.
* Updated documentation.
* Updated dependencies.
* Removed support for python 3.6.

-------------------

* Retrained all models with scikit-learn 0.24.2
* The Portuguese model has improved and is now also at 99%+ accuracy.
* Updated documentation.
* Updated dependencies.

-------------------

* Unknown verb inference is now faster.
* Added more tests to the test suite.
* Automated setup.py dependency injection.
* Updated documentation.
* Updated dependencies.

-------------------

* | Added new `GitHub Actions Workflows`_ to automatically build, install and test mlconjug3
| on Windows-x64, MacOs-x64 and Linux-x64 for maximum copmatibilty.
| The package should build and install even on the newest MacBook with an Apple ARM M1 processor
| by installing mlconjug3 through conda_ instead of pip:

.. code-block:: console

$ conda install -c conda-forge mlconjug3

* Updated the documentation to make it clearer.
* Updated development dependencies.