Inspiration
The "Breaking News" hack from the 2015 Numenta Spring Hackathon
What it does
The system monitors semantic changes in a text stream, in this case in the Twitter feeds of several US Presidential candidates. The application displays anomaly scores in the stream calculated by the HTM, giving the user an indication of unusual topic changes detected in the stream. The user can then inspect the actual text in the stream using the UI to investigate the anomalies found.
How we built it
We used the Twitter API to retrieve Tweets for six top presidential candidates. After aggregating the Tweets by day, we create a semantic SDR representation for each day's Tweets using the Cortical.io API. These SDRs are then input into an HTM for each candidate, which learns the semantic patterns contained in candidate's posts and computes anomaly scores. We built a frontend in JavaScript that communicates with the HTM via a REST API implemented in Python to display the data to the user in an interactive web application.
Built With
- javascript
- nupic
- python
Log in or sign up for Devpost to join the conversation.