twss 0.1

A Naive Bayes classifier that can identify double entendres.

TWSS
====

This is an implementation of a simple double entendre classifier in Python.

This currently uses a Naive Bayes classifier (the NLTK implementation) as a
Python package. This was inspired by the `bvandenvos Ruby TWSS project
<https://github.com/bvandenbos/twss/>`_ and uses the same data corpus.

This was built on the eve of `Barcamp Mumbai <http://barcampmumbai.org/>`_ 8
and presented during a session there.

Suggestions welcome. Do file bugs. Fork away. Send us pull requests.

Setup Instructions
------------------

.. code-block:: bash
$ virtualenv --no-site-packages --distribute venv
$ source venv/bin/activate
$ pip install -r requirements.txt

This creates a virtual environment for this project and install all the
packages necessary for the project to work.

Demo
----

Once this is installed, you can take it out for a spin:

.. code-block:: python

>>> from twss import TWSS
>>> twss = TWSS()
>>> twss("That was hard")
True
>>> twss("Hello world")
False

The first call can take a while- the module needs to train the classifier
against the pre-installed training dataset.

Getting dirty
-------------

You can supply your own training data using positive and negative corpus files:

.. code-block:: python

>>> twss = TWSS(positive_corpus_file=open('foo.txt'), negative_corpus_file=open('bar.txt'))

or directly, as a list of tuples:

.. code-block:: python

>>> training_data = [
... ("Sentence 1", True),
... ("Sentence 2", False),
...
... ]
>>> twss = TWSS(training_data)

Roadmap
-------

- Making this pip-installable.
- Writing a sample web app.
- Writing a sample Twitter client.