27

In order to help people find duplicate questions as they ask, we are now showing Similar Questions in the right hand sidebar on the /ask page.

Similar Questions

This is deployed to all sites right now. Here's how it works:

  • it is specific to the /ask page
  • you must have entered a title (any length) and a minimum body length of 220 characters.
  • every 45 seconds while you compose your question, we will query for Similar Questions based on the title, body, and tags you have entered
  • if we found a different set of similar questions since the last time we checked, the sidebar on the right will update with a fade animation and the Similar Questions, replacing the How to Format help.

You can mouse over the similar question links to get a tooltip preview of the question bodies.

(Yes, we do already show Related Questions after you leave the title field, but that is based purely on a title-based match and does not look at body or tags at all.)

This is currently experimental, but the goal is to help question askers find potential duplicates and related questions without bugging them with interstitial pages. Test it out, preferably when asking real questions (as I am now!)

If you have feedback on this, let us know.

8
  • Hopefully the number of duplicate questions will significantly decrease :) May 17, 2011 at 1:28
  • 1
    it seemed to take a while to come up for me. does it not come up until you've entered tags? because by the time the user is picking tags they're probably just about to post, likely won't stick around for another 45 seconds
    – Kip
    May 17, 2011 at 2:39
  • @kip if you read the post above, it says minimum 220 chars plus 45 second interval. So the first time it will appear is between 1 and 45 seconds after you type the 220th character in the body. May 17, 2011 at 3:05
  • @Jeff: Should the 220 chars kick in if you've copy and pasted rather than typed it in? May 17, 2011 at 3:07
  • 4
    Seems a little confusing that both "related" and "similar" questions are being shown. How is the typical user to know the difference between the algorithms used to match each type? Why not just combine them all in a single list, filtering as more data is collected? May 17, 2011 at 8:27
  • @cody one happens much faster than the other; see my comment on Renesis' answer May 17, 2011 at 10:53
  • @JeffAtwood: re "one happens faster than the other": why not go ahead and do the fast one, and then update the list as you get more information?
    – Kip
    May 17, 2011 at 13:39
  • @kip just search here on meta for the unbelievable love-fest over title-only matching. I think this is ridiculous, but if it works for people then I don't want to interfere with that. May 17, 2011 at 13:44

3 Answers 3

11

This is a great idea, I was thinking about this just today when I saw Add "from the Related sidebar" to the auto-generated "Possible duplicate" comment, since it seems like the problem there is that those questions weren't showing up before the "Post your Question" button was hit. (and, given how many different feature-requests there were for it.)

However, the only feedback I have, is that there will probably be a bit of user confusion over "Related Questions" and "Similar Questions" both showing up at the same time: "They sound like the same thing — but show different results. I'm confused. And sad."

(Proposal - title change and maybe some Javascript attention on when related questions shows/hides)

1
  • 7
    I was thinking this too; I checked in a change to make the title search say more clearly what it actually is: Questions with similar titles May 17, 2011 at 10:51
7

Fantastic idea! My feedback:

On my screen the bottom edge of the box is clipped which looks a bit broken (this box scrolls with the window)

Screenshot of feedback box clipping

My screen resolution is 1366x768 (laptop) and I'm using Firefox 4.0.1

Also I think it would be useful to reduce or eliminate the character limit - many simple questions can be expressed perfectly adequately in less than 220 characters and now that I know this is here I'll probably start entering my tags before the question body to see if I get any matches.

For example my test question was:

I can't find any iron!

I want to build myself a large mine cart system in Minecraft but I can't find enough iron!

What can I do to try and find more iron?

[Minecraft] [iron]

Once I copied and pasted the content of the above question a couple of times however the first result was dead on the money How to get a lot of iron in Minecraft? - in contrast this question was only on the 2nd or 3rd "page" of the Related Questions box so this is definitely helping.

10
  • +1 for the feedback. The same is happening to me :( May 17, 2011 at 7:57
  • 1
    "I think it would be useful to reduce or eliminate the character limit" you should know that questions below 220 characters in length get penalized in various ways. So.. no. May 17, 2011 at 9:15
  • @Jeff: I have the impression that you're shooting yourself in the foot. You're right, very short questions tend to be (but do not have to be) bad questions. Nevertheless, they're are still allowed, and by not showing the "Similar Questions" you make them even worse! (Even if the majority of short-question-askers ignores the "Similar Questions", there are legitimate short questions and thoughtful askers that do care.) May 17, 2011 at 10:43
  • 1
    @hendrik I find your opinions fascinating, but I have data that tells me otherwise. May 17, 2011 at 10:45
  • 2
    @Jeff: It would be fascinating to learn more about that data :-) May 17, 2011 at 10:46
  • @hendrik download the data dumps and go nuts -- but just realize that deleted content is not there. And 99.99% of our worst content is deleted. You can diff the data dumps to see what got deleted between them. If you are thinking "wow, that sounds like a lot of work" well, that's because it is. :) May 17, 2011 at 10:49
  • 1
    @Jeff: Thanks for the feedback. I still don't quite see the point you're making. Are you saying that short questions are always bad? Or that showing "Similar Questions" wouldn't help anyway? Or is it that showing the "Similar Questions" is too expensive? (The latter I'd understand: If it helps only for a fraction of the short questions, and it's expensive, then everything is clear.) May 17, 2011 at 10:53
  • @Jeff I don't understand your reasoning either. What good does withholding the new "Similar questions" list (which I think is great) from askers of (short|bad) questions do? Is it because the algorithm would catch and list only similar (short|bad) questions?
    – Pekka
    May 17, 2011 at 11:01
  • 2
    @pekka there's not enough to search on -- test it yourself rather than asking me. Besides that, users who continually ask short questions may find "interesting" things happening to their accounts.. automatically. Just sayin'. May 17, 2011 at 11:02
  • @Jeff: I see you once wrote "tags is the big one". If this is still true, then I don't understand "there's not enough to search on". (After all, "Related" is also shown for short questions, isn't it?) Regarding your "Besides": I agree with you 100% on "users who continually ask short questions", but I'm not talking about those. I'm talking only about thoughtful askers with occasional short questions. Why don't you help them avoid duplicates? May 17, 2011 at 12:01
1

TL;DR

When tags or title are not focused, the 'Similar Questions' should be refreshed and displayed by default.

Current Behavior

When you edit the title, you are given this side bar:

Title Sidebar

When you edit the tags, you are given this side bar:

Tags Sidebar

When you edit the body, you get the similar questions side bar:

Related Posts

When you defocus title or tags, you keep whatever side bar you had. So if you were editing title, you keep info on what the title should be (even though you are no longer editing the title), and ditto with tags.

Issue

Tags can dramatically change the contents of the similar questions. If a user writes a question without any features identifying which language it is (writing a question about jquery not mentioning jquery), the sidebar questions will be very wrong until I enter tags.

Here is a very poor question which demonstrates the behavior:

How do I map multiple items together in to a CSV?

I want to use .map() to combine multiple items in to a CSV (comma-delimited) output.

For instance, if I use this code, I can get a list of all <div> elements in the SO main page:

$(#question-summary > div).map().group().join()

However, what if I want to get the info for question score, views, number of answers, and question title on a single line of the CSV (an element of the array produced by the above code)?

This gives me the following sidebar focusing on CSV:

Similar Questions with No Tags

But as soon as I add the jquery tag, I get much more relevant results:

Similar Questions with the JQuery tag

Suggestion

Since tags are usually edited last, and can have a significant impact on the quality of the results, it would be much more user-friendly if losing focus on title or tags would by default show the Similar Questions (refreshed of course) as if the user had clicked back on the body again, since this is the most important info.

You must log in to answer this question.

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