tweetcoder

• Alexander Biggs
• Francisco Canas
• Sasa Milics

What be this?

A very work-in-progress twitter trends visualization device using autoencoders.

What does this need?

• NLTK python lib -- www.nltk.org Install with: pip install nltk

• Theano -- http://deeplearning.net/software/theano/ I recommend following the easy installation instructions here for your particular brand of operating system: http://deeplearning.net/software/theano/install.html#install Note that Theano has its own set of dependencies as outlined in the install guide linked to above, including:

• NumPy -- http://www.numpy.org/

• SciPy -- http://scipy.org/

What are these files?

src/jtotext.py -- Script used to produce plain text, line-return separated list of tweet text from a json file dumped by the db we are using.

src/preprocessory.py -- Script used to extract feature vectors from a file containing the body text from tweets. Will use two separate and mutually exclusive methods: "Bag of Words" and "per Character" feature extraction.

src/autoencoder/autoencoder.py -- The main code file for the autoencoder. This file will contain the code that trains RBMs, links them together, and encodes input vectors into tinier input vectors.

src/autoencoder/deeplearning/rbm.py -- The Restricted Boltzmann Machine class that is a building block for the autoencoder.