Some videos are posted under different titles on the same source site for videos. We have several source sites that have the same videos. Some videos may vary by a few seconds as well. Duplication is going to happen.

Yeah, I know it's not an exact science. That's why I'm suggesting it needs to be quick to search variations of title and tag.

It already does this when you hit submit, it offers you five possible duplicates before you can hit the final submit.

How would you suggest improving this further?

Also an exact dupe of an embed tag already throws an error.

winkler, you could just search for the video before you start copy & pasting. I find the new submission works pretty well, "When I paste the embed code, it could know if it's a dupe." -> it does! Obviously there are 18 versions of the same video on YouTube and they should merge them all.. but also the pseudo-match by tags & title show five top hits, which works for most videos. Unless the original or your video submission is way off on description, I find it works pretty good.

So I guess you are suggesting AJAX submission page to get more immediate feedback, in case you never bothered to search.

Please try http://instantdomainsearch.com/ , and enter some text to find a domain name.

This exemplifies a quick search.

James,
On the submit page, maybe it's a waste of time if exact dupes are rare but I'm imagining, for exact dupes:
1. Click submit video
2. Enter embed code
3. DHTML fires to check for dupes. Puts up a DUPE symbol.

So you want it to do it on the submit page rather than making you go to the next page? I mean it does already locate exact dupes of the embed tag. I guess this is something that we would maybe eventually implement, but currently since it's already pretty well covered I would say that it's pretty low on our development list.

When I paste the embed code, it could know if it's a dupe.
Reading the description of your problem with the submitting system, I can understand that you can be irked after fancifully crafting a snappy title and some witty tags, but there honestly isn't much value in implementing such a thing simply to check for duplicate embed code. If your creativity is really that important to you, there's nothing stopping you from just clicking the 'continue' button immediately after pasting the embed code. If it's a dupe, the error page will tell you immediately.

Beyond that, I'm imagining something which would return an HTML table with video name and tags as I search. I don't need all the other metadata (submitter,date,votes,rank) to know if it's a dupe.
I'm fuzzy on this point. You said "as I search," but are you saying you'd like a dynamic search/display for titles/tags as you type your new video's title/tags? If this is what you mean, the effort just seem like it'd be worth it. You're saying just title/tags is enough to be positive there's a dupe, but if we show a table of titles/tags, strictly that by itself (not browsing the actual vids) just cannot be enough to know there are dupes in 99.9% of cases. Most titles/tags are not that obvious; you would surely still have to go look at the actual videos. Of course, the alternative is what we do today: just click 'continue' and see an actual listing of the equivalent results. I know the number of concurrent submitters at any one moment isn't vast, but the overhead involved in performing a search through all the thousands of rows in the video db for every keystroke is just a bit much for too little return.

Don't get me wrong. I'm not at all against suggestions or ideas of how to improve things here on the Sift. I'm an enormous proponent of new features, but in this case I think there just aren't strong enough arguments in their favor. Of course, that's just one man's opinion. (Oh, I mean mine, if that wasn't obvious. )

I guess the issue I'm thinking about is this - I've had a couple posts get past the dupe detection, so it's not perfect. This dupe post is an example
- it didn't find the orginal.
I had searched for dominos too.. I thought this was a new killer entry. I see the problem now- dominos vs. dominoes.

I assume the dupe detection works is additive - the more tags match, the higher the ranking. So, searching for an indidual tag (dominos), even Ajax-style, has less value than a compound key search (tag1..tagN,title)

Guess I'll do a workaround on my part- do a quick pass on tags and title, check for dupes, then wordsmith if necessary. Thanks for the discussion.

That is the biggest problem with anything automated: the human part. When people misspell (sometimes typo'ing with the precision of a sledgehammer) it doesn't do well to help the rest of us. (But, hey, we can always use the nifty 'edit tags' feature to try aiding those with such literary deficiencies, right? .

Fyi, you're right about the levels of ranking when matching keywords; the more matching terms, the closer to the top of the search results.

winkler, that's why our second level of sifting is humans. people have pretty good memory, so when we see dominoes again, i think i duped it and fixed the original tags because of spelling problem. In this case, no AJAX would help because I had to google for it, but I had the advantage of knowing it was in here somewhere. So it does work, and this is a good example of how a good sifter enables this site to be cool. (pats self on back).. but seriously, don't worry about lost votes or dupes. I still post a bunch, and others are there to smite down my videos.. because they happened to have seen them the first time around.