Hi! I'm currently attending Wikimania (I have a session on Friday at 4.30pm).
Tilman Bayer suggested to share this tool and techniques here, so I am following his advice :).
I've been using Google BigQuery for a while to analyze Wikipedia's publicly available data. It's main advantages:
- It's unbelievable fast (try it - operations that you might expect to run in minutes or hours run in seconds). - It's secure, but you can also instantly share data (no need to download and setup locally before being able to analyze - BigQuery is always on). - Everyone can use BigQuery with a free quota of 1TB of monthly analysis.
Interesting links:
- Quick getting started: https://www.reddit.com/r/bigquery/comments/3dg9le/analyzing_50_billion_wikip... - Analyzing the gender gap in Wikipedia (Freebase, and joining it with pageviews): https://www.youtube.com/watch?v=lV5vk3higvA - Massive Geo-Ip geolocation from the changelog: https://www.reddit.com/r/bigquery/comments/1zh7ty/massive_geoip_geolocation_... - Just for fun, the most popular numbers: https://www.reddit.com/r/bigquery/comments/2p0vz4/query_of_the_day_the_most_... - Top Wikipedia Entries Which Are Most-Edited by Members of the U.S. Congress http://minimaxir.com/2014/07/caucus-needed/ - Music recommendations: http://apassant.net/2014/07/11/music-recommendations-300m-data-points-sql/
I have a couple other interesting examples I haven't written about, but the invitation here is for you to try your own :).
My main challenge today: How to get more publicly available data into BigQuery. Let's work together :). I'm sitting around the big data analytics team today at the Wikimedia hackathon - and as said earlier, I'll do a session on this topic on Friday at 4:30pm.
Thanks!
Thanks Felipe! Yes, I think this is a really interesting tool to explore.
Another quick example: List articles attract between 2-3% of pageviews on the English Wikipedia:
SELECT SUM(requests) FROM [fh-bigquery:wikipedia.pagecounts_20150715_14] WHERE LEFT(TITLE, 8) = 'List_of_' AND language = 'en'
244091
SELECT SUM(requests) FROM [fh-bigquery:wikipedia.pagecounts_20150715_14] WHERE language = 'en'
8870277
(Caveats: during one hour this Wednesday, using the old pageview definition, i.e. not excluding spiders and bots, and relying on the article name instead of categories.)
I understand Felipe has already been talking to the WMF Analytics team, who are making major progress on https://phabricator.wikimedia.org/T44259 currently.
On Thu, Jul 16, 2015 at 11:32 AM, Felipe Hoffa felipe.hoffa@gmail.com wrote:
Hi! I'm currently attending Wikimania (I have a session on Friday at 4.30pm).
Tilman Bayer suggested to share this tool and techniques here, so I am following his advice :).
I've been using Google BigQuery for a while to analyze Wikipedia's publicly available data. It's main advantages:
- It's unbelievable fast (try it - operations that you might expect to run
in minutes or hours run in seconds).
- It's secure, but you can also instantly share data (no need to download
and setup locally before being able to analyze - BigQuery is always on).
- Everyone can use BigQuery with a free quota of 1TB of monthly analysis.
Interesting links:
- Quick getting started:
https://www.reddit.com/r/bigquery/comments/3dg9le/analyzing_50_billion_wikip...
- Analyzing the gender gap in Wikipedia (Freebase, and joining it with
pageviews): https://www.youtube.com/watch?v=lV5vk3higvA
- Massive Geo-Ip geolocation from the changelog:
https://www.reddit.com/r/bigquery/comments/1zh7ty/massive_geoip_geolocation_...
- Just for fun, the most popular numbers:
https://www.reddit.com/r/bigquery/comments/2p0vz4/query_of_the_day_the_most_...
- Top Wikipedia Entries Which Are Most-Edited by Members of the U.S.
Congress http://minimaxir.com/2014/07/caucus-needed/
- Music recommendations:
http://apassant.net/2014/07/11/music-recommendations-300m-data-points-sql/
I have a couple other interesting examples I haven't written about, but the invitation here is for you to try your own :).
My main challenge today: How to get more publicly available data into BigQuery. Let's work together :). I'm sitting around the big data analytics team today at the Wikimedia hackathon - and as said earlier, I'll do a session on this topic on Friday at 4:30pm.
Thanks!
Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l
wiki-research-l@lists.wikimedia.org