> And if you need to prepend a + to query terms, that means for the given query terms, the missing word is rarely occurring and the overall ranking is low
Isn't that the point of searching: find a specific thing in a whole bag of other things? I don't use Google to find the Yahoo homepage (I know many people actually do; should we help them???)
If you were right, Altavista would have prevailed and Google would not have emerged; the way Altavista worked was, they linked the words in a query with the "OR" operator, so that any query produced at least some results, however irrelevant.
If you were right, Google would have done this from the start; why are they doing it now?
(Could it be because focus groups are taking over?)
Besides, what I'm complaining about is different from automatic spell correction; when you mistype a query Google automatically searches for (what it thinks are) the correctly spelled words, but it tells you about it and let you search for your original query with just one click.
Having "add" match "append" is done silently, without any visible option to disable it.
And if you have many words that you need to prefix with + it gets crazy; in theory there's the "allintext:" parameter, but if you use it often you get flagged as a bot and can't search at all.
I think I'd pay to have access to a "Google Pro" service where all these tricks would be disabled; no Instant, no fuzzy search, no nothing. Like 2001 Google.
> Isn't that the point of searching: find a specific thing in a whole bag of other things?
Web search isn't as simple as you make out to be. It's not a 'select * from blah where field like %like%'. Searching is about finding relevant information. Keywords are taken into account, but again, it's not a giant hashtable.
If I search for 'python file system watch', it shows me results for 'monitor' as well. Most of the cases, the keywords are the means to find the end result.
When you enter the queryterms, they pass through a lot of filters - removal of stopwords, n-gram tagging, spell check, hyponyms/hypernyms translations, lemmatization...before hitting the index. And then the ranking itself is quite complex.
And it's simply that sometimes 1 or 2 out of n query terms doesn't make it to the first page.
I don't work for Google and I don't know about how Google does it, but the notion that Google is ORing query terms is absurd. It just can't be. If Google were ORing query terms, you would never find MIT if you searched for "massachusetts institute of technology" or vice-versa.
> If you were right, Google would have done this from the start; why are they doing it now?
Once again, they aren't ORing query terms. The missing terms are falling out of the front page due to low rank.
> I don't use Google to find the Yahoo homepage (I know many people actually do; should we help them???)
Nothing special needs to be done for them. The generic ranking works great.
> If you were right, Altavista would have prevailed and Google would not have emerged; the way Altavista worked was, they linked the words in a query with the "OR" operator, so that any query produced at least some results, however irrelevant.
Google prevailed because of better ranking algorithm.
> (Could it be because focus groups are taking over?)
Focus group doesn't decide how algorithmic search works.
Search engine internals are really not my point here (I build Lucene-based search engines for a living).
My point is about user experience and user expectations.
I expect a search engine to find documents that contain the words I'm searching for, not other words it thinks are equivalent, or some other document that has a better "ranking". Ranking is sorting the pages that actually contain words in the query, not in abstracto. Ranking is about relevance, not popularity; popularity is but one piece of information when calculating relevance.
> The missing terms are falling out of the front page due to low rank
In the example I gave, a word is replaced by another word, not just ignored; one can tell because the replaced word is highlighted in the search results, as if it was the one I was looking for.
> Focus group doesn't decide how algorithmic search works
Focus groups can help decide if a feature is or isn't worth it. Some user feedback is useful; relying too much on user feedback and losing sight of the original mission of the company is dangerous. (Remember "Organize the world's information" and all that?)
- - -
It's likely there are solid and sane reasons why Google is doing this (and of course they're free to do as they want -- since we don't pay for their services who are we to complain); what I don't get is why these "features" cannot be turned off (as can Instant).
I don't understand why you're getting downvoted. I agree completely, and have had the same thoughts many times myself. Especially about longing for a Pro type of service where we could get back to actual, effective search.
Isn't that the point of searching: find a specific thing in a whole bag of other things? I don't use Google to find the Yahoo homepage (I know many people actually do; should we help them???)
If you were right, Altavista would have prevailed and Google would not have emerged; the way Altavista worked was, they linked the words in a query with the "OR" operator, so that any query produced at least some results, however irrelevant.
If you were right, Google would have done this from the start; why are they doing it now?
(Could it be because focus groups are taking over?)
Besides, what I'm complaining about is different from automatic spell correction; when you mistype a query Google automatically searches for (what it thinks are) the correctly spelled words, but it tells you about it and let you search for your original query with just one click.
Having "add" match "append" is done silently, without any visible option to disable it.
And if you have many words that you need to prefix with + it gets crazy; in theory there's the "allintext:" parameter, but if you use it often you get flagged as a bot and can't search at all.
I think I'd pay to have access to a "Google Pro" service where all these tricks would be disabled; no Instant, no fuzzy search, no nothing. Like 2001 Google.