- TwitHunter - a twitter client with scoring
- permanent link
- April 24, 2009
-
Motivation
I follow one hundred persons on Twitter, and keeping up gets increasingly time consuming.
I could just unfollow some of them but I don't want to because sometimes I actually have enough time to read them.
Also, fun or interesting twits may be written by people posting tons of junk twits besides.
Scoring
I envisionned a scoring system, where you define rules that increase or decrease the score of individual tweets.
That way, when you only have 5 minutes to spend on 5 hours of tweets, you just read the tweets with the highest scores.
Implementation
TwitHunter is an attempt to see how scoring could be useful in a Twitter client. (Using the ugliest Twitter client on earth is not a part of the experiment.)
So far, it is just a quick hack (ca 8 hours work), but it works :-)
Basically every tweet gets 50 points. The score is then changed according to simple rules, per user or per keyword.
For instance, on the screenshot, Sebastien's tweet has 50+15 points for mentioning iPhone (keyword rule) and Fraser's one got 50+10 (user rule).
The slider is set on 53, so only tweets with 53 points or more are displayed.
TwitHunter is build on top of Matt Gemmel's MGTwitterEngine.
Disclaimer: the data model is subject to change at anytime, so don't rely on it to store your data for now.
Project
The code is here: http://github.com/nst/twithunter/
I have neigher time nor interest to write a "real" Cocoa Twitter client.
Instead, I would like the scoring approach, if considered useful, to be added to full featured clients. Maybe someone will take it into Canari?
I would also like to try out baysian filtering when I have time.
So, don't hesitate to fork the project, it is still very young, I only added 1000 lines of my code and some Cocoa bindings, so it is still very easy to change.
By the way, my Twitter account is @nst021.
