7

In another question, I expressed my distress over editing a post only to be told at the end of the process that someone else have a more substantive edit and therefore my edit cannot be accepted. In that previous question, I took the algorithm as a given while trying to raise a discussion on the matter.

Well one fatal flaw in this algorithm is that if I just type a bunch of junk on the keyboard like

fsaddsafdsasafdsa;ljkfgdsahgk;jkdsa;fljka;s;lkfa;ljksafd;ljkasf;ljkdsaljk;fda;

this registers as a better edit and will be accepted. Can we take a fresh look at what is considered the better edit and improve the algorithm so it is more fair to the 2 or more people trying to edit at the same time? Of course, not trying to game the system is always a good start.

  • 1
    Do you have any suggestions on how it would be done without just comparing the quantity of changes? Determining the quality of changes is impossible for a computer, so we have to trust that the 2K user who made the biggest change would likely have made the best one. – psubsee2003 Feb 5 '14 at 12:53
  • So you suggest to invent algorithm that parse both edits, learn what each means and decide which is more correct based on some super advanced AI? – Shadow The Princess Wizard Feb 5 '14 at 12:53
  • Well, we can check against an English dictionary as a start. Beyond that, yes I would have to think about what other things could be done. Maybe does the edit relate to the topic which would involve getting into some NLP I guess. – demongolem Feb 5 '14 at 12:58
  • @psubsee2003 Actually, I don't think it is all that bad to use reputation as part of the equation. If you have earned more trust, there is a greater chance that we will accept your edit as better – demongolem Feb 5 '14 at 12:59
  • @demongolem but then there are a lot of cases of names of classes, code, object names are not going to match English words, so how will that work? And rep is a factor already. Edits by 2K users will always override suggested edits. – psubsee2003 Feb 5 '14 at 13:01
  • @psubsee Yes I suppose the code is the tricky part. And then if we let code go, I just put my above statement inside a code block and beat the algorithm. We would have to resort to accepting the tagging as correct, check to see if the code is syntactically valid and then really hope it does the right thing. But I see that we do have spam filters already in place, can we apply these to edits? – demongolem Feb 5 '14 at 13:04
  • Using the English dictionary seems bound to fail. For one, users make typos - and that doesn't necessarily make it a bad edit (baby, bath water, and all that). There is also plenty of jargon used in software development, that isn't found in the dictionary, both in the code and when describing it (my favorite is performant). Finally, 2K users that are trying to game the system by mashing their keyboard with their palm are not going to make it very far... is this really a problem that has happened often? Examples? – Aaron Bertrand Feb 5 '14 at 20:54
1

How about some sort of scoring system that factored in things like reputation of each user, words that relate to the topic (or the tags of the post), words that are in the English dictionary, previously accepted edits made by each user, and finally, number of changes. Whichever edit scores higher is accepted. And in the event of a tie or two close scores, a mod could review them and resolve discrepancies.

  • Might also consider number of previous accepted edits by the user editing. – DaImTo Feb 5 '14 at 13:55
  • 1
    That would be an excellent idea, answer edited. Thanks! – Winderps Feb 5 '14 at 13:57
  • no software is perfect. such system would just increase to load of moderation required. – user221081 Feb 5 '14 at 14:05
  • You're right it would, but when it comes down to it there's a trade-off that's necessary in order to keep the highest quality possible throughout the site, and I think a system like this would end up with a happy compromise between moderation and such content. And the moderation would only come into play when two users' scores end up very close to or exactly the same. – Winderps Feb 5 '14 at 15:03
  • @DaImTo that only works for people < 2K (whose edits go through review anyway). For those >= 2K, at what point do the rejected edits from before they hit 2K become irrelevant? What if they never made any edits at all before they hit 2K (I believe that is true for me, for example, not that my pre-2K history would really be indicative of my editing skills today anyway). – Aaron Bertrand Feb 5 '14 at 20:51
1

I agree with you it would be nice to improve the algorithm. Meanwhile, if I feel that my edit is really going to improve a post I select all the text in the box and copy it before moving on. Then see the current edit and kind of compare the changes. If it seems alright then just move on but if I feel that my changes are better than edit again simply pasting my work and overriding the whole thing.

  • I do the exact same thing. Usually I end up making a second edit. I give up, though, when the author is robo-editing (like changing one letter at a time). That gets annoying pretty quickly. – Aaron Bertrand Feb 5 '14 at 20:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .