Blog

RSS
  • Fuzzy Link Bot

    14 Abr 2010, 0:18

    Fuzzy Link Bot - http://atomboy.isa-geek.com/flb/ is a Flex app I wrote to visualize the relationships between people, places, and bands that are mentioned in music news articles. I have a script that checks a bunch of RSS feeds every hour. It downloads the text of articles, tries to strip out the ads and navigational fluff to get just the content, and sends the creamy center of every article to the Calais (http://www.opencalais.com/) service. Calais sends back its best guess at who what where was mentioned in the article and my script stuffs the results into a database.

    If two things are mentioned near each other, they might be connected. If they are mentioned together often, the strength of the connection is greater. Fuzzy Link Bot uses a formula to get the percentage of how often one thing appears near another, and takes into account the number of different ways they are mentioned to get the top few links for an entity.

    There's more technical details at the site. It's not quite finished, but I think it's as close as it's going to get for this summer anyway. Comments most welcome!

    Cheers,
    Andrew