Check out the ipython notebooks for examples.
- network_search.py
- message_classification.py
- twitter_stream_classification.py
Explores the messages, locations and connection of a user account of interest. Information is stored in an sqlite database created for the source_node. One potential risk is the exponential growth for each network depth. The user can counteract this in two ways: i) Retrieve information from only the top x% connections and ii) limit search to a specified network depth.
Trains a text classifier on labeled examples. Then classifies tweets stored in an sqlite database to identify intersting candidates for further inspection.
Identify interesting candidate tweets from the (keyword filtered) twitter stream. Use a trained text classifier to keep only tweets which fall into classes of interest with a minimal prediction probability.