MeFi Search for Exact Phrases Using Speech Marks?

November 21, 2002 2:07 PM

Mefi search functionality and double posts. Using google extensively I've developed the habit of searching for phrases by using speech marks in order to retrieve more specific results. Metafilter search doesn't seem to support this feature, and partly as a result I made a double post, after searching for "moon trees" instead of moon trees. I'll know for next time, but could more double posts result from other new users with the same habit? Possibly.
posted by nthdegx to Feature Requests at 2:07 PM (24 comments total)

Feature requests seemed as good a category as any, although I would stress I am not making a judgement call on this issue, but merely asking the question.

And in case it sounds like I am trying to remove responsibility from myself for making a double post, some points:

1. had I had a look at the resulting search url I made have been given a clue that my search hadn't gone entirely according to plan.

2. due to the size of mefi's archives compared to the amount of data google trawls, even if "searching with speech marks" was supported, it would have been much safer for me to search for the words separately than as a single phrase. There would never be many posts with both the words moon and tree, whether they were double posts or not.

That said, I think anything that could reduce the instances of double posts is a good thing, especially given the usual response.

Also - if this is a non-issue because of some stupid factor I have overlooked please point it out to me (once is enough, though).

Ta!
posted by nthdegx at 2:13 PM on November 21, 2002

Wow, a preemptive call-out defense.
posted by Yelling At Nothing at 2:31 PM on November 21, 2002

Best way...
posted by nthdegx at 2:40 PM on November 21, 2002

This request has come up before, and my answer is that I'm not much of a search engine programmer, and I've spent a few hours trying to add boolean capabilities to the search and it's a bit beyond my reach (I understand the required SQL, and what needs to be done, the implementation never worked -- it's most likely syntax problems with the ColdFusion and/or SQL code). So the search engine is a bit basic.
posted by mathowie (staff) at 2:43 PM on November 21, 2002

parsing the input and whacking together the query SEEMS straightforward at first blush. then one tries to do it and enters a world of hurt. the sheer number of input permutations to be dealt with is mind boggling :-)
posted by quonsar at 2:48 PM on November 21, 2002

Worth a comment or two above the search box, then? Seems a reasonable solution.
posted by nthdegx at 2:56 PM on November 21, 2002

Worth a comment or two above the search box, then? Seems a reasonable solution.

I usually don't expect search engines to support things like that unless they explicitly say so (which they normally do). But I guess it's worth a warning.
posted by mathowie (staff) at 3:00 PM on November 21, 2002

I pretty much assume search engines support searchstrings within quotes unless told otherwise, either explicitly or by the age of the page. And I usually forget to leave the quotes off MeFi searches, so a note would be good.
posted by me3dia at 3:11 PM on November 21, 2002

if mefi used a seperate database for search keywords, where thread numbers are associated to keywords, it could probably be pretty easy to do boolean stuff. but we've had that discussion before, and the problems are still the same: there's no time to do the programming, and no likelihood to open the source of mefi for others to do so.

google, i should say, has never supported search strings. there was never a need, for its algorithm returns results such that those who have the keywords close together in the source text are ranked ahead of those who have the keywords far apart.
posted by moz at 3:18 PM on November 21, 2002

(generally speaking, i should say, google ranks those near in proximity higher: for there are other factors considered as well.)
posted by moz at 3:21 PM on November 21, 2002

I pretty much assume search engines support searchstrings within quotes unless told otherwise, either explicitly or by the age of the page

I think there's a distinction between real search engines and joe-blow's weblog.

I assume that the search page at excite, the IRS, or Amazon will support some advanced features without saying so (though they usually explain it on their advanced search pages). When I'm looking on someone's blog, through their Atomz or Movable Type search engine, I don't assume anything more than a simple brute force search (but then I've made rudimentary search engines so I'm aware of what the baseline features are).
posted by mathowie (staff) at 3:25 PM on November 21, 2002

In my case it was purely force of habit. So used to googling I really didn't give it any thought whatsoever. No one's fault but my own but I'm just thinking about minimising the risk of other's doing the same thing.
posted by nthdegx at 3:29 PM on November 21, 2002

google, i should say, has never supported search strings. there was never a need, for its algorithm returns results such that those who have the keywords close together in the source text are ranked ahead of those who have the keywords far apart

That's not been my experience. If I do a search for just the words, and then do another search with the words enclosed in quotes, I get distinctly different results.

For instance, compare the results on the following two vanity searches:

with quotes | without quotes
posted by willnot at 4:33 PM on November 21, 2002

strike that - I just noticed that the two different searches I linked are hitting different indexes which may account for the differences I saw.
posted by willnot at 4:34 PM on November 21, 2002

Or maybe the indexes aren't different and it's just that the scope text is contextually different. I'm going to crawl back to work now.
posted by willnot at 4:36 PM on November 21, 2002

google, i should say, has never supported search strings.

Wrong. Enclosing a phrase in quotations results in an exact search, particularly in cases where any of said words are the "common" words google will otherwise ignore. Compare a search for who likes pancakes versus one for "who likes pancakes".

Similarly, the top ten (of 11,300) results for rabbit pancake are quite different from the top ten (of 15) for "rabbit pancake". One would assume that, if quotes were meaningless, the results would have identical relevance, but that doesn't appear to be the case.
posted by Danelope at 6:43 PM on November 21, 2002

One would assume that, if quotes were meaningless, the results would have identical relevance, but that doesn't appear to be the case.

sorry; my mistake.
posted by moz at 7:32 PM on November 21, 2002

(i should explain: years ago, when google started, i recall someone saying that on account of google's search heuristics, exact quote searches could not be used. i've carried that for all those years, never questioning. but, hey: we all make mistakes.)
posted by moz at 7:35 PM on November 21, 2002

No worries, moz. If anything, your search results will be more accurate now. Heh.
posted by Danelope at 7:46 PM on November 21, 2002

From Gogle's Advanced Search Tips:

Phrase Searches

Search for complete phrases by enclosing them in quotation marks. Words enclosed in double quotes ("like this") will appear together in all results exactly as you have entered them. Phrase searches are especially useful when searching for famous sayings or proper names.
posted by DBAPaul at 5:59 AM on November 22, 2002

Another way to do a phrase search is joining the dots. I thought this was subtly different to a phrase search, but cannot now construct any examples that are. Could have sworn it implied an ordering of the joined words in the search result, but didn't necessarily look for exact matches. But it looks like I was wrong ...

What I like to do is to use phrase searches in conjunction with or opposition to another keyword. Thus "joining.the.dots +puzzle" is quite different to "joining.the.dots -puzzle". And yes, google is still my command line interface to the web, despite flirting with teoma from time to time for a different set of results.
posted by walrus at 7:10 AM on November 22, 2002

Use site specific search. add site:metafilter.com to the end of your query. example. Good night.
posted by blue_beetle at 9:52 AM on November 22, 2002

Easiest solution for quote marks on MeFi might be to simplify the search string by removing punctuation before doing the query. Or else search for both the original search term, plus a simplified one containing only alphanumeric characters.
posted by kindall at 10:22 AM on November 22, 2002

blue_beetle - as accurate as your ideal solution is, it does not resolve the problem of the casual searcher / excited poster, and as such does not address the issue at all...

i agree with kindall... some sort of forced simplification of the searched query would be ideal. mefi's "catalogue" of data would be small enough in the majority of cases for the searcher to spot no difference at all.
posted by nthdegx at 5:49 PM on November 22, 2002