Tweetement is a web-service that tries to improve search experience on Twitter. We take the results retrieved by the original Twitter Search query submitted via its REST API, text mine these results, perform query expansion via pseudo-relevance feedback with new terms, and display the most interesting conversations that are happening now.

Live: http://tweetement.com/ (or http://tweetement0.appspot.com/)

Demo: http://demo.tweetement.com/ (or http://demo.tweetement0.appspot.com/)

Project details: http://uaf57924.ddns.uark.edu/?page_id=249 and http://bit.ly/kb-thesis

• Python 2.7 and pip
• Google App Engine (GAE) SDK for Python

git clone git@github.com:Bekt/tweetement.git
cd tweetement
pip install -t lib/pip/ -r requirements.txt

• You need to have a Twitter app registered. Once you have a registerd app, follow instructions in credentials.txt.

At this point, everything should be ready to go. To run the web service locally:

# From project root.
<path-to-gae-sdk>/dev_appserver.py .

The service should be running at http://localhost:8080

You also should seed the list of stop words into your local datastore. You only need to perform this once. Go to http://localhost:8000/console and execute:

import seeds

seeds.seed_stopwords()

Make sure the entries have been entered at: http://localhost:8000/datastore

• The frontend is written in AngularJS + Bootstrap.
• See templates/ and assets/.

High overview of the flow:

• User submits a query via enqueue().
• enqueue() creates a Query model and puts it in a queue (background job).
• The Query is popped from the queue (tqueue.pop()), ExpandedQuery is created and query expansion is performed (tqueue.expand_query()) by calling Twitter Search API multiple times.