best include every letter May 13, 2010 11:06 AM   Subscribe

Search peculiarity: MeFi searches not picking up sub-terms?

I searched for "pixel" before posting my now deleted post, and found nothing. HuronBob searched for "pixels" and found the previously posted thread. Shouldn't search be more effective in situations like this?

(Not grousing that I was deleted... I just seem to hit against this wall and have posts deleted more than I should, because I DO search before I post.)
posted by hippybear to Bugs at 11:06 AM (19 comments total)

It's MetaFilter, not Google. If you want stemming in your search, Google is the way to go. Prior to Google, when I searched for something on AltaVista, most of my searches had some enormous clauses like (tv OR television OR teevee OR telly).
posted by adipocere at 11:11 AM on May 13, 2010


Right... except that stemming seems to work fine on keywords... So we know that technology is built into the MeFi search on some level...
posted by hippybear at 11:16 AM on May 13, 2010


Sorry hippybear, not sure why stemming is working in some cases and not others. That's something we can look into. Another path for checking duplicates is looking through a couple of common tags.

I think doubles are just going to happen in some cases no matter how many checks you do. In this case the YouTube video ID changed so the link-double-checker didn't work either. There were just multiple places that slipped through the technology cracks here.
posted by pb (staff) at 11:19 AM on May 13, 2010


I have no idea how pb has done it, but I know that Google's stemming, aside from some front-loaded just as a seed, comes from the size of the corpus. I guess what I'm getting at is that Google has quite a few more programmers than MetaFilter does, with a lot more data to play with. Google-quality search ain't easy.
posted by adipocere at 11:20 AM on May 13, 2010


YouTube changes IDs? Crap. I know some folks who are going to be really unhappy about that.
posted by adipocere at 11:21 AM on May 13, 2010


Thanks for looking into it, pb. I really don't mind being deleted, as I know it's part of the game. I'd just like to have a better shot at not having it happen. And yes, I'm a bit lazy with my keywords -- I know now to eschew using roots and search on variations of terms, I guess.

At least now I have some bit of understanding why I've had (what I feel are) too many deletions in the past while -- I'm searching but not playing by the rules of the search engine!
posted by hippybear at 11:23 AM on May 13, 2010


I have no idea how pb has done it...

It's not me, it's Microsoft. We use SQL Server Full-Text search and it handles the stemming. I can make sure we're taking advantage of that feature as much as possible, but the internals are a black box to me.
posted by pb (staff) at 11:23 AM on May 13, 2010


YouTube changes IDs?

Um... no. The original thread's link was "removed due to a copyright claim" or something. I had linked to the official post by the production company which created the video, which obviously didn't have the same YouTube ID as the post which was taken down.
posted by hippybear at 11:24 AM on May 13, 2010


YouTube changes IDs?

Well, videos get deleted and that ID goes away. Then someone uploads the same video and it has a different ID. So the IDs don't change per se. They just move around.
posted by pb (staff) at 11:26 AM on May 13, 2010


Ah. Okay. I've used that and one of its cousins in projects before. Urgh. I was at a search engine conference not long ago and we had a big discussion on Why Search Is So Freakin' Hard. Amongst all of the others was that the bar just keeps moving up.
posted by adipocere at 11:28 AM on May 13, 2010


Yeah, Google makes search seem effortless and they shape all of our expectations. We offer links to the Google Search equivalent of our internal search on every results page here, and that's another way to double-check what you get.

Changing our internal search engine is in our long-term plans. But no search service is as perfect as Google—there will always be issues.
posted by pb (staff) at 11:34 AM on May 13, 2010


It's not me, it's Microsoft.





begins pouring bronze into casting bucket
posted by infini at 12:04 PM on May 13, 2010


But no search service is as perfect as Google
Bing
posted by tellurian at 12:18 PM on May 13, 2010


Bing? You reach for a counter-argument and you bring up Bing?

Bing is Ned Ryerson: dorky, unpolished, expects you to love him instantly, and blatantly trying to sell you stuff.

Fah!
posted by adipocere at 12:32 PM on May 13, 2010


Tag search is the shit.
posted by cashman at 12:45 PM on May 13, 2010


Bing John Malkovich
posted by cortex (staff) at 12:45 PM on May 13, 2010


If google was as good as cuil, a search for 'pixel' would definitely return results for sea cucumbers.
posted by Rhomboid at 12:58 PM on May 13, 2010


shush, he was trying to say Ping... its the accent


btw, must say that Google has been crappy of late... ridiculous results, major differences between different country extensions and why no image results?
posted by infini at 1:00 PM on May 13, 2010


Why I don't trust bing.

I mean, seriously.

(in case they fix this later: the first non-sponsored hit is "Why are Mac's So Expensive? - Yahoo! Answers")
posted by idiopath at 2:40 PM on May 13, 2010 [1 favorite]


« Older Can we talk about it?   |   San Diego meetup! Newer »

You are not logged in, either login or create an account to post comments