filtering: March 2006 Archives

Drop in just five new lines of code and you got a noise keyword filter. Some keywords are “noise”, not adding any significant value being there. So now we have a filter that removes common English words such as “that”, “this”, “with” from the list, giving more focus to the actual “content” that people put there. And we can add and remove keywords with a few mouseclicks with no need to hardcode anything anywhere. I like automatic admin backend generation :)

UPDATE Feb 2007: Someone asked what’s the noise keyword list. There’s no online list, but at this time, the noise keywords are: “than”, “that”, “from”, “this”, “with”.

We now filter “too similar” moods. If you ping over a mood that is slightly different from the previous few ones to this user, but still “too similar”, you will get a “200 Too Similar” response (also updated the pinger protocol post to reflect that) and it won’t be captured. This is to make the database more interesting and representative and less biased towards overloading it with only slight changes.