Ask HN: Review my app: Firetweets

Post by timhastings » Sun, Aug 01, 2010, 10:41 PM UTC

iampims wrote: post #20

timhastings wrote: post #18

Earlier quoted context omitted.

Thank you. It already works for any hashtags or usernames that it encounters, here's #sundaybaconclub http://tagwalk.com/tag/sundaybaconclub

I'm impressed. From what I've read on your blog, you have gathered a massive amount of data. How do you fight spam?

Post by timhastings » Sun, Aug 01, 2010, 11:17 PM UTC

iampims wrote: post #22

timhastings wrote: post #21

Earlier quoted context omitted.

It's many pronged approach. User timelines are loaded for "reputable" users, (see blog for computing reputation) and search input is used with primative bayesian filtering. In my experience, I have found Tweet source (web, Tweetdeck, API etc.) to be one of the most effective quality/authenticity indicators.

Using the tweet source is such a smart idea! So far I rely on username filtering / patterns. Toughest is to get rid of job postings ! Good luck with Tagwalk ! You've got a new user ;)