Research talk:MDM - The Magical Difference Machine

I suppose this is not still in progress? Did it have any outcome? Maybe EpochFail knows. --Nemo 12:36, 11 March 2014 (UTC)[reply]

Hey Nemo. This happened. See dumps and code code. Sadly, we haven't re-run the diff generation since 2011. On the bright side, I've been working towards getting a similar system up and running again in my spare time. I was actually just sitting down to write down a specification for an updated, pure-json diff format. If you're interested in helping out, my next problem is block move detection and I could use a study buddy :) --EpochFail (talk)

Hm, I'm not sure I should embark on yet another project. How about Flöck and Acosta [1], I've read it and it's quite interesting. I didn't manage to properly digest it though. --Nemo 10:43, 1 June 2014 (UTC)[reply]

Hey Nemo. I've generalized Flöck & Acosta's strategy into a "Segment matcher" that will work for arbitrary segmentation strategies (e.g. the paragraphs and sentences used by Flöck & Acosta). I'm putting together a system that will allow for live diff synchronization with a MediaWiki installation. If you'd like to track my progress or check out the segment matcher, see https://github.com/halfak/Difference-Engine. --EpochFail (talk) 23:12, 19 June 2014 (UTC)[reply]

Bah! I forgot to explain *why* I generalized the WikiWho algorithm. I figure we'll want to use a similar strategy for non-western languages that don't break nicely into paragraphs and sentence. :) I'll need some help writing a parser for those though. --EpochFail (talk) 23:15, 19 June 2014 (UTC)[reply]