By ResearchBuzz
I’ve made v2 of Gossip Machine and it’s about 1000% better!
I made the first version of Gossip Machine last summer. It uses Wikipedia pageview data to find days when Wikipedia articles got especially busy traffic. Those dates are then turned into single-day Google News searches.
Using past indicators of interest is an useful way to find meaningful search results about a news subject, but I was never entirely happy with the first version of Gossip Machine. The math used to calculate popularity is sloppy, there’s no indicator in the result list which days are the most busy, and it analyzes a year at a time, which doesn’t allow for closer assessment.
Curly and I kicked these problems around and made Gossip Machine v2. WOW I am pleased.
This version analyzes only a month of page views at a time. For each day it generates a z-score. Z-scores above 1 are filtered and presented in order of Z-score, with a red bar indicating how busy the page was compared to the mean for the month.
I LOVE how this works. It’s lightning fast and the red bar provides a visual indicator that’s understandable at a glance. And the Google News results are still generally good, though if the topic is too general or the page counts are too low it gives wonky results.
Now I gotta figure out how to make use of those negative z-scores…
June 12, 2023 at 07:18PM
via ResearchBuzz https://ift.tt/ZPeV3Tq
No comments:
Post a Comment