Monday, January 20, 2014

HN Sentiment

HN Sentiment

Check it out: HN Sentiment

Web application that allows users to anylize the sentiment of the Hacker News community towards a particular topic

Built using:          

This porject was done over a few days of my Christmas '13 holidays, for the Iron.io hackathon (where it placed 2nd). The intent is to generalize the way the entire HN community feels towards a particular topic.

Technical Stuff

This project is all open-sourced under the The MIT License and available here

The project is written in CoffeeScript on NodeJS using the Express framework. How it functions:

  1. The application generates http requests for the 10 (number in flux) most recent posts with the entered keyword in either their content or title.
  2. The content is then parsed to determine the original text and, more importantly, all of the comments.
  3. The app then aggregates title, content, and all of the comments of the post and analyzes them for any sentiment using the AFINN-111 wordlist to determine an integer value for the desired topic.
  4. The number discovered is then normalized by the number of posts found with comments then returned and displayed to the client.

In my next iteration I will be:

  • Work on better caching (right now have fairly basic cacheing, which on Heroku gets dumped every time the repo sleeps).
  • Track down any bottlenecks or bugs

Send me an email with any suggestions!!

No comments:

Post a Comment