Google adds verbatim search mode for your exact search terms | Hacker News Reader

Google adds verbatim search mode for your exact search terms

insidesearch.blogspot.com

263 pointsMatt_Cutts162 comments

I know that a lot of people were sad that we switched from +word to "word" to do an exact match for a word.

The good news is that we're rolling out a "literal mode" that will search for exactly the words you typed. It does a verbatim match with your words, so it turns off things like spelling corrections, stemming, synonyms, optionalized terms, and so on. I think we saw multiple proposals for this on Hacker News, so thanks for the suggestion.

Matt, do you know if there are any plans to once again allow the mixing of exact match with the regular fuzzier search functionality? It's great to have exact matching back, but I often found it a useful strategy to build the search loosely, then refine with +specifics. Perhaps something like this, in exact match mode:

  ~run ~sprint ~hurdle nike

could be the equivalent of the old:

  run sprint hurdle +nike

Yes please allow something like this, mixing verbatim with fuzzier search functionality is powerful.

You still can do that, you just have to use "quotes" "around" "the" "words" instead of +putting +the +plus +sign +before +each +word. This just lets you get that type of search for every word easier.

Isn't

    run sprint hurdle "nike"

what you want?

Yes, and of course that works fine. It's just the ease of using an inline prefix to modify the term, rather than enclosing it in quotes. We're mourning the loss of the +.

Thank you. Irritation has been growing that I can't search for what I want without fielding all kinds of "well what about this variant..." results, without even + prefix and "" squishing into some unwanted twist. Would be nice to make (at least as an option) Verbatim mode the default, with a link provided per search to show variants.

This isn't really the solution we were looking for Matt.

Can we replace the + operator with something else? How about¬? We don't use that very often to the point I don't even know what it's called, but it's there on the keyboard in one keypress.

Not on US keyboards, I don't think.

Depends on your OS. ¬ is a quick alt+l on any Mac. However, it's the symbol for logical negation[1], and thus probably not a great substitute for +.

[1]http://en.wikipedia.org/wiki/%C2%AC

On Windows, you have to do Alt+0172 to get a ¬ unless there's some shortcut I don't know offhand.

I've looked all over my German keyboard, with its crazy-assed bracket locations, and I don't see it anywhere.

My UK keyboard has it left of the '1' key above the qwerty keys

I have carat and a little circle. The bastards.

Speak for yourself. I've been wanting this solution for years.

I'll pass the feedback on. Part of the issue with + is that it turned out a lot of people were mistyping queries with + when as far as we could tell, they weren't trying to match an exact word. Getting further off the beaten path with a more rare character might help though.

Exclamation point? (!this !word !exactly, !darnit!) Or a double exclamation point if that's still too common...

It really was nice to be able to control which words were verbatim and which weren't. Quotes are perfectly logical but just a pain to retrofit onto a query you're refining.

Surely simpler soultion would have been to respond to a search for "+this is a +search" with:

Showing results for this is a search Search instead for +this is a +search

As google does for other "typos".

Failing that, a replacement for + would be welcome. I can't imagine many people type, for example ^word or =word - allowing use of a symbol that indicates promotion of that word word in the search terms, they way + did.

The difficulty will be reeducating the (presumably quite large number of - even if proportionally a small percentage of googles user base) people who have been using + in searches for the last 10-15 years (I'm almost positive this concept existed in other search engines before google) that + isn't + any more, and please use this other symbol instead.

Currently Google offers me the ability to change the background image on the Google home page.

How about the ability to use something like Google "classic" with the old way of using the + sign?

Hide it as an off-home page selection, that way only people who really want it will choose it.

I'd also be interested in something similar for an "advanced search syntax" mode. Probably not very useful for the majority of people, but could be something a minority might be very passionate about.

What about the equals sign?

=exactstring

Westlaw search syntax would be awesome!!!

That's the symbol for logical negation, which would make it a confusing replacement for +.

> This isn't really the solution we were looking for

What? Yes, yes it is. Please don't take it away now!

That character is most often used as a logical negation operator, so it may be confusing to use it as a force-inclusion operator for search.

It's also almost never found on standard US keyboards, which is why the tilde is often used as a logical negation operator in it's place.

finally... and thanks.

Out of curiosity, how does google judge search result quality? It seems as though search is trending toward the idea that more results are better, when usually the opposite is true.

A bit speculative: this is more related to Ad revenue. A wider spectrum of search results means more Ad variety and the ability to serve more advertisers per search query.

I'm annoyed by Google's search quality by the day. I'm starting to use DuckDuckGo more (unless there are other alternatives).

I know it's tempting but please don't speculate like that, it raises my blood pressure. :)

Google keeps ranking changes completely siloed from the ad side of the company, and the decision makers about ranking changes don't see statistics about how it might effect ads. We don't even have a way of collecting that data. (At least as far as I've ever seen.)

I couldn't resist the speculation temptation :)

Correct me if I'm wrong; When I'm searching for (unquoted) "Canadian equivalent to SOPA Bill". I'm getting results about Soap, Telenovela, Spas, Dishwasher Soap, TV shows (and 1 related hit). The result, as you mentioned, could be ranked independent of Ads, but the Ad spectrum is wider (Soaps, Spas, Soap operas, TV shows, etc...) [I'm not getting any in this particular case].

If it's restrictive search (quoting each word - sigh!), the search result quality is slightly better (first 3 hits), but the potential of Ads is also low.

I know you put the user first - but wouldn't you speculate in my case?

I'm not sure that logic works. You're assuming that the Ads are matched to the query in the same way that results are matched to the query - that a more restrictive query necessarily translates into fewer ad matches.

There's another possibility which is that the ads are the query and the search query is in fact their match result. In that case, adding or removing quotes within the search query might not have any impact on the ads being shown.

Don't know which way it works, but it seems to me the latter would be better for showing ads without having any negative impact on search results.

Ads has a very sophisticated system of matching ad topics against queries. That's how they make the big money. :) I work in search though, so I don't really know anything about how it works except at an extremely high level. I'd be really surprised if that sort of effect actually happened though.

Moultano: I 100% believe google keeps ranking changes siloed from the ad side, but Google 100% favors its own properties within its search results so you can't be surprised that someone thinks this way? Here are some examples for you that need major changes in my opinion.

www.google.com/#hl=en&q=san+francisco+restaurant+reviews www.google.com/#hl=en&q=hollywood+cauldron

-when i previously used google for retail reviews I would just bypass all of your results to get to yelp. sometimes trip advisor.

-look at those youtube video results for the 2nd query. So off the mark, but you guys keep them in there because it is good for google. it doesn't help your users and makes you look bad. I don't know what you guys are thinking.

I use Bing for 95% of my searches now. For tech searches I still think you guys are better, but for everything else I use Bing.

Here's an official post on the subject: http://googleblog.blogspot.com/2008/09/search-evaluation-at-...

Why aren't you allowing for both; The original +prefix as well as the quotes?

Although "sad" with where Google search is (was) going isn't the right term, I am now very much unpissed, so many many thanks for this.

How is "literal mode" different from what "allintext:" promised? And is there an option to turn this on permanently?

I always used the + to mean "the following word or quote surrounded string HAS to appear in the result text".

Was this incorrect?

I seem to recall reading in prior discussion on this that +foo included stemming while "foo" does not? (And, therefor, some people were unhappy with "foo" not just due to the syntax change but also due to the difference in behavior.)

Some other people in this thread have suggested ¬ as a substitute (and no, it's not on a typical U.S. keyboard). Others point out that ¬ means negation in a typical context. Do you folks index backticks? If not, what about backtick? I guess that wouldn't be too *NIXy; however, most keyboards seem to have one, and if it's not currently indexed, might be a choice. (It's also "sort of quotish", in nature.)

P.S. Hmm... Backtick could cause unexpected results in the case of someone pasting a single quoted value from Word, or similar. Might not be the common, though, and I never did like Word's "smart quotes", anyway. ;-)

I'm not seeing the change yet, will we be able to use "verbatim:hello world" to do an exact search? I don't see myself navigating through the sidebar each time I need to find something specific.

Yes, please, verbatim: or similar would be great, especially for use on a mobile.

This is a good idea, it will allow us to use the string interpolation address bar search functionality we all love in our favourite browsers.

In general, nice one google for once again keeping an ear to the ground.

Still waiting on being able to search for punctuation. Being able to search for "$. ruby" or "$_ perl" or "$: << '.' ruby" would be super handy.

edit: Just realized "$_ perl" actually does work. It would be nice, however, if we could search for these kinds of terms generally and not just in specific case.

1000 times this. I have been waiting a decade to be able to search for '1/2' or '6' 3"' or '£3.50' (not necessarily numbers, you get the point i'm sure), and i was dissapointed to find this post wasn't announcing that.

I launched a search engine that does exactly this just over a week ago, you can see it here: http://symbolhound.com/

I freaked out when I saw this headline and thought it meant google had finally decided to implement this feature a week after my launch.

That's freaking fantastic. Shame it didn't get much notice on proggit. I'll resubmit it at a busier time and share it on G+ and whatnot. You might want to submit it here yourself, so you can answer any questions people might have and let the community get to know you. Try submitting it with the "Show HN:" heading. Send me an email (in my profile) when you do submit it and I'll be sure to swing by and give it one of the critical first upvotes.

Really cool tool; thanks for that.

Thanks! I do plan to post it on HN, I just want to wait until it's a bit better (right now stackoverflow dominates the results, so I need to adjust the ranking system and index more sites).

BTW, I don't see your email in your profile.

>BTW, I don't see your email in your profile.

Sorry about that. I would have thought putting it in the email field would have made it visible. I've put it in the about field now.

Great! Nice, simple and easy. It can even search for negative numbers without interpreting the - as "exclude". A literal search engine, without Google's well-meaning "intelligence" can be useful in many cases.

This. Half my kingdom for this feature.

One of the biggest problems I have with Google search at the moment is when I need to look up specific functions or symbol names. If I'm looking for documentation on g_find_str, the LAST thing I need is to see results that contain "... are going. Find string is a ..."

That's great, Google does n-grams/symbol search finally. Tho 'f# "|>" ' is way less common than "$_ perl", doesn't work yet.

And i think the folks that did, e.g. hoogle took some time to get it right for 1 language

http://www.haskell.org/hoogle/

http://scalex.org/

Nice, now search personalization can be turned off!

> With the verbatim tool on, we’ll use the literal words you entered without making normal improvements such as <...> personalizing your search by using information such as sites you’ve visited before

Personally, I appreciate this more than allowing to search for exact matches. Hopefully ‘filter bubble’ just became less… bubbley. =)

Can you confirm or discuss more about what this does to the "filter bubble?"

In two windows of the same browser, I typed gyros, one immediately offered me an instant search result of gyros Phoenix (where I currrently am), the other window, in verbatim mode offered me gyros, gyroscope, and gyros recipe.

(Don't ask me why I tried gyros as my search example.)

I believe verbatim mode turns off personalization, and it looks like it turns off localization as part of that. So the query [yoga] shows local places to do yoga for me, but in verbatim mode I don't see that.

Thanks Matt. This helps with at least one of my complaints about the direction Google search has been heading.

(Now if you'd just eliminate that godawful slow URL redirector on result links…)

Why does this query not do what one would expect.

    halo -"halo 2" -"halo 3"

Also, removing "Advanced Search" from the main google.com and from the top of a search results page beats out moving "cached" to the preview page in terms of "stupidest decision Google has ever made".

I'll ask about the halo query. In the mean time, it looks like [halo -halo.2 -halo.3] does what you want.

Do you see a gear symbol in the top of main google.com and the results page? I think that's where Advanced Search is now.

Thanks for the query recommendation. Curious why the other didn't work, I feel it (or similar ones) have worked as I might expect in the past.

>Do you see a gear symbol in the top of main google.com and the results page? I think that's where Advanced Search is now.

Seems Android action-bar-esque almost, but yes, thanks.

Could you guys consider bringing out a case-sensitive search feature sometime?

And, of course, the "search for what I actually fucking typed" mode is completely undiscoverable, hidden behind another click, and which can't be defaulted into.

This is a good bandaid, but is still a bandaid. I have still switched away from Google to a search engine that will actually search for what I told it to, not one that thinks it can guess my meaning. (Not to mention one that doesn't wrap all the outgoing URLs in click-tracking garbage then uses JavaScript in an attempt to hide it, with no way to opt-out.)

For the vast majority of users, things like spelling correction are a huge help. Roughly 10% of all queries are misspelled. Switching the default to "search for exactly what I typed" would hurt those users and lead to more spam and malware for the average user.

We try to find the balance that helps users the most, and searching for a term with "term" or the literal mode provides an escape hatch for power users.

If only Google had some sort of "account" system that could store user-specific preferences...

It'd be better if we could just append a "v" after the URL to make verbatim searches. That way we can bookmark it and there will be no need to be logged in.

You can. Append:

  &tbs=li:1

Ugly, but it works.

Heh, I just found this too and added it to my default omnibox search in Chrome.

You may have to add your own to the list:

hXXp://www.google.com/search?q=%s&le=en&start=0&tbs=li:1

How do you tell the difference (for mis-spelled words) between people doing an actual search and people spell checking a word? It could be that a significant percentage of the 10% are not "search queries" but "spelling queries"

Personally, I like the previous "did you mean" behavior; if Google thinks I misspelled a word, or that I want a synonym, by all means give me the alternative, but by default I want the exact query I searched for, because more often than not I really did mean what I searched for rather than what Google thought I meant.

Can't agree enough this. I'm pretty sure user interface research has shown that it's incredibly frustrating to users when you tell a computer to do one thing and it does something else. The suggestions are helpful, but it should default to what I tell the fucking thing to search for!

If the computer successfully guesses exactly the thing you wanted and does that instead of what you said, it can make you marginally more happy than a "did you mean", since it avoids the reaction of "well why didn't you just do it then?".

However, if the computer guesses wrong, that will cause significantly more frustration to the user. This disproportionate amount of annoyance means that the computer must guess correctly far more often than it guesses wrong. Unfortunately, Google's current system frequently guesses wrong, and then compounds those incorrect guesses by using them by default, taunting the user with their original search.

That's been our observation as well, and we've tried to constrain launches in exactly that way. Techniques that can fail in a way that disrespects the query have to have a much much higher win to loss ratio in order to launch.

What I think we're understanding now is that something that fails 1% of the time, might not fail for just 1% of queries, it might fail for 1% of the users and work perfectly for the rest, and that small set can be really unhappy as a result. This is one small step in the direction of fixing that, but it's an issue we're paying close attention to, thanks in part to the feedback we've seen on HN.

We very much want to avoid systematic failures like this, and if technical queries are a blind spot then please send them over whenever you see them. My email is in my profile and I'll make sure they get routed appropriately, and I tend to watch threads about Google trying to get good examples to debug.

I'm getting worse and worse results for all sorts of queries over the past few years, not just technical ones. I can't think of specific examples from the main search right now (I finally snapped and switched to DDG a few weeks ago), but here's a particularly terrible one from the Android Market:

https://market.android.com/search?q=yore&so=1&c=apps

I appreciate that Google thinks I'm an illiterate chimp who doesn't know how to spell "your", but really truly, I was looking for a game with "yore" in the title. That's why I typed it. It's even an English word.

> What I think we're understanding now is that something that fails 1% of the time, might not fail for just 1% of queries, it might fail for 1% of the users and work perfectly for the rest, and that small set can be really unhappy as a result.

Excellent insight. Linux kernel developers have had a similar experience with the use of the likely() and unlikely() hints on conditionals; it took a while to realize that unlikely() has to mean unlikely for any user, not just "unlikely for most users, and likely for the rest". So, for example, it doesn't make sense to write "if (unlikely(feature_disabled))", because anyone with the feature disabled will meet that condition every time.

I've definitely had the experience that Google's automatic correction does the wrong thing for me more often than not; I almost always end up clicking the link to go to the search results I actually asked for.

Send over the examples when you come across them (or if you have it turned on, try www.google.com/searchhistory/ to find queries you've issued in the past.) I'll make sure they get to the right place.

Actually, spelling might be the only exception to that rule.

If you genuinely misspell a word, it's nice to have it fix it for you in one motion.

Funny, for me spelling is least likely to be an exception to the rule. If I misspelled a word in my query then I meant it to be misspelled ... I was probably trying to track down a specific article in which I knew the word to be misspelled, or else I wanted to know how common the misspelling was. And my experience is that Google hates me a little more every year.

Meanwhile, the synonym replacement is actually helpful sometimes. I wish I could turn off one without the other: that's why "verbatim mode" isn't going to do it for me.

"I'm pretty sure user interface research has shown that it's incredibly frustrating to users when you tell a computer to do one thing and it does something else."

You honestly believe that Google is doing something that is worse for the vast majority of their users? That makes the vast majority of their users more frustrated? And that they know this is true because research has proved it?

Why would they act like that?

Most of the people in this thread keep forgetting that they're not regular users. I'm willing to bet that the average Google user (90% of the population) often mistypes. I know I do, and I'm a pretty good speller and typist.

You keep forgetting that most people can't type fast, can't spell that well, and you know what, a lot of the people typing on Google don't even know English all that well, either. For them, Google's spelling auto-correct is incredibly useful.

You're putting words in my mouth. Please don't do that.

I think it should be a given that it's frustrating when a computer doesn't do what you tell it. That doesn't mean that the cost of that frustration isn't offset by the added convenience for many users. In my case it isn't, and those (relatively frequent) times when it does frustrate me, I seriously consider switching search engines.

"You're putting words in my mouth. Please don't do that."

You're right. I'm sorry.

For the record, most of my comment was addressed to everyone in this thread, not just to you.

There's a magic "nfpr=1" flag in the query string that seems to revert back to the old "did you mean" behaviour. Sometimes it randomly fails though. I don't know if this feature is going away with this update.

Fully agree with this. Approximately 90% of the time Google thinks I misspelled something, it's wrong.

For me it's the opposite but I'd love to be able to manually toggle these things on or off.

My ideal setup might be to get url bar searches to show regular google and ctrl-k searches to be google verbatim searches.

This is especially true on Android and other low bandwidth/hard to type devices, where I have time and again searched for the specific exact thing I am looking for only to have instant search and spelling correction help me find something I could not care less about.

I guess I'm in the minority here. I'll even often use Google as my spell checker.

I think the algorithm does too many "magic" things lately. I've noticed this when I had some search queries for "flask" (python micro web framework) and some programming related input and most of the returned results were about "flash".

We're always trying to find the right balance. The nice thing about this change is that it gives people an escape hatch if we get it wrong. In theory, I could imagine we could look at the corresponding data to improve the balance in the future too.

Could you not add a hacker mode? Put it at google.com/h or something and remove all the cuddly fuzziness that drives us power users completely mad. I'm sure your intentions are good and that the latest changes are helpful for many or most users. But I for one have always enjoyed Googles assumption that the user is intelligent and therefore never autocorrected me, countrary to for instance MS office. Suggesting corrections does not interrupt my flow and are generally helpful. Auto-corrections are in great part highly disruptive for me. Please don't deviate from this very sound principle, at least not for us power users.

I think of this literal mode as pretty close to the hacker mode. Most of the time our query rewriting is an improvement, but we wanted it to be very quick (just a couple clicks) to go into hacker mode if we get things wrong.

My hunch is that someone could a search shortcut for verbatim mode, but if you used it enough you'd find that you missed the query improvements.

Can I enable literal mode permanently with a cookie? Otherwise, you better make sure that going into literal mode is fewer clicks than, say, going to DuckDuckGo.

I was thinking the same thing lately. I've wondering if I was crazy to notice that Google was ignoring my " ".

It is quite infuriating.

Spelling correction has nothing to do with "" being worse than +.

I think a lot of us would welcome spelling suggestions while we're using our + operator.

For that 10% figure to be meaningful, we'd need to know how many people are trying to search for something precise and not receiving it.

Although now is not the time to be picky (to be thankful, rather), why not simply let logged in users default to literal search, while still displaying the "did you mean" suggestions somewhere on the page...?

I use Google as a spellcheck quite a bit, because it's generally better at working out what my tortured attempt at spelling really means than dictionary sites / apps.

I wonder how many of the 10% are intentional.

All I really want is for google to tell me when it didn't search for what I wrote. I put a word in the search, I expect it in the results - even if the result has a low page rank.

If you want to default considering some words optional, that's fine. But say that above the search results.

What search engine do you use now, out of curiosity?

Probably DuckDuckGo.

Yes, DDG. My only complaint is that it is slower than Google, but it's worth it to me to get the results for what I actually typed the first time.

> wrap all the outgoing URLs in click-tracking garbage then uses JavaScript in an attempt to hide it, with no way to opt-out

Speaking of which, is there any way to disable or bypass that kind of crap? (e.g. with a user-script or something?) It's incredibly obnoxious. And sadly Google's not the only on that does it.

Frankly, it seems like a major design mistake from my perspective that Javascript is allowed to manipulate the status bar at all. It'd be nice to be able to remove that ability, if not "un-wrap" the links entirely.

> Speaking of which, is there any way to disable or bypass that kind of crap?

If you have JavaScript disabled, or these particular event handlers somehow blocked, the href is never replaced.

> Javascript is allowed to manipulate the status bar

That's disabled by default in most browsers (see in FF about:config, dom.disable_window_status_change -> true). Look at the mousedown handler of links to see that the href attribute is changed on mousedown.

The problem is blocking google.com javascript will also block recaptcha on any domains that use it (a lot do), which they ever so helpfully moved from recaptcha.org to google.com.

The href attribute is still changed. While javascript can't change the status bar explicitly any more, they can set the link's href to the fake URL and change the link's target in the click event.

Demo: http://jsfiddle.net/zTRQh/

There are a variety of scripts that remove the link tracking. I use this one: http://userscripts.org/scripts/show/95092

I think it's forgotten in this discussion that the HN community is probably very different from the larger internet population when it comes to search queries. We all tend to do a lot more "exact/verbatim type searches" such as looking up some library package, whereas your regular user has a lot more "soft" queries where they're looking for a concept / business / song lyric rather than an exact phrase, and having proactive spelling correction is a huge help.

Correcting spelling automatically also probably helps guide those users towards higher quality search results rather than squatters who have purposely targeted misspellings of popular search terms to suck in naive searchers.

It's not realistic to expect a search engine targeted at all users to tailor the UX to this group; the best you can expect is some secondary options or commands that allow the extra control we want.

Counter example:

I just tried googling [how pronounce resig]. It thinks I want [how pronounce resignation]. Both +resig and "resig" didn't work (heck, I even tried +"resig"). Had to add -resign -resignation.

There is nothing technical about this search (unless Resig is such a unique last name that belongs to only John Resig). I'm sure there's lots more counter examples. I really doubt normal non-technical people actually likes the new google search.

Btw, what's the answer? Only link I found was [1] someone says it's REH-sig, not REE-sig (not too sure what's the difference). But I'm more interested on how to pronounce the second part, is it more like "side"("sign" minus 'n'), "seed", "sad", or something else?

[1] http://www.stevesouders.com/blog/2009/09/22/sau-drs-sour-not...

[how to pronounce "resig"] works for me now. (Or at least doesn't mess with the query, though the results aren't great.) Do you remember what results you were seeing before?

yea, that seems to work now. Wasn't getting any "resig"'s at all before unless I added -resign -resignation. Now, ["pronounce" "resig"] seems to do the trick, as w/o quotes or just '+'s still gives resign and resignation results.

Actually, I don't think that's accurate. It's not auto-correcting "resig" to "resignation" as if you misspelled it. It just naturally happens that "pronounce" and "resig" almost always appear together when "resig" is a part of "resignation". Evidence for this is that your post is now hit number seven.

I'd suggest that if there's 100 users entering the query [how to pronounce resig], there are a lot more of them that have dropped the "n" from resign than are looking for the pronunciation of resig. And even more queries overall where "resig" was supposed to be "resign". I admit it's unfortunate that sites have to guess based on the behavior of the group (rather than knowing, for example, that you usually don't misspell things) but I'm not sure any of us individually has enough query history to make those kinds of guesses.

Thanks for also considering the power users/edge cases.

Even better would be to have a prefix for the search, e.g. /l(iteral) so that you could do the whole search by keyboard only (but perhaps there is a key binding for the menu option? Would be great)

Hooray! Does the "more search tools" menu learn from the user which tools are useful to them? I seem to recall at one point having to open the search tools for the helpful timeframe narrowing options, but now those routinely appear exposed on the left.

If after a few times using verbatim it realized I was a "power user" and always had that option sitting on the side for me, I'd love it.

This is excellent news. I know when Google does stuff that they base it on evidence, and they do it for the best of most users, and that my edge cases are not what most users would have.

I was getting caught in some frustrating searches, so I'm really pleased about this.

There's some amazing figures in the article:

> However, we found that users typed the “+” operator in less than half a percent of all searches, and two thirds of the time, it was used incorrectly.

Wow.

Finally: If Google offered user controllable stemming that'd be awesome. When I started using Google there was no stemming. Now there's either Google suggested stemming or user-controlled no stemming.

> we’re also applying similar ideas directly to our algorithms, such as tuning the accuracy of when our query broadening search improvements trigger

As you are doing this, please keep in mind that when you incorrectly broaden my query, and there is no immediately obvious way to re-narrow it, it costs you much more than you gain each time you correctly broaden my query.

Does anybody else feel somewhat uncomfortable to live in a world where people who a) know how to spell and b) actually want a search engine to search what they are looking for are considered "edge cases/power users"?

I've always laughed at the idea of the internet and search engines making people dumber, but this seems like hard evidence. Most Google users are too stupid to have ever been able to use a library...

> actually want a search engine to search what they are looking for are considered "edge cases/power users"?

Arguably synonyms and word stemming still are what they are looking for. You could easily argue that one either way.

Word stemming is often desirable. I wish it were possible to turn on just that, and only for certain words.

Well, I feel uncomfortable using a product that is designed to fill such people's needs at the expense of mine.

Just like Yahoo's homepage full of entertainment news, Google's aggressive query correction sends me the message "You are not our target audience, because there isn't enough money in catering to you".

And when Google is sending that message to geeks, I can't help feeling there's some brand destruction going on.

In the past, we provided users with the “+” operator to help you search for specific terms. However, we found that users typed the “+” operator in less than half a percent of all searches, and two thirds of the time, it was used incorrectly. A couple of weeks ago we removed the “+” operator, encouraging the use of the double quotes, which are more likely to be used correctly.

Actually, Google decided to repurpose the + operator for Google Plus Pages Direct Connect (http://www.google.com/support/plus/bin/answer.py?hl=en&p...).

that may have had something to do with it, but i think you can probably trust the official sources that it really wasn't used much and was often used incorrectly. i've met more than a few old people who had been told at some point that they always needed to put plus signs between words instead of spaces, no matter what.

I hope they also add a way to enable it from within the query string itself, like 'verbatim: foo'. Much easier, and that way you can do it from search boxes and such too.

Hi Matt,

The problem is that, sometimes, even quoted queries don't guarantee that the word will be there.

I hope verbatim solves that.

Yes, I believe it does. I believe verbatim requires an on-page page.

Thank you so much for this. I know I'm one of the less-than-a-percentage-point of users for whom the suggestions were making me cranky; this will make me a much happier Google user.

Hey Matt, thanks for listening.

Since code search is gone, I'd really love a “grep” for the Web. Full page search, including headers,js,css,html comments, etc.

If you could pass that along, I believe it would make a lot of us geeks happy.

This sounds like it has potential, although I can't test it yet. You should try searching Google for information on COM programming. Try finding information on "COM Events", for example. Good luck. Macy's is on the the first page! Bing is not any better.

I get "Understanding COM Event Handling" at #1, but agree that it's not ideal. Some words like COM or IT have dual meanings where sometimes you want to treat them like stopwords and sometimes you don't.

Just tried it with verbatim but results are still no good =( I can see how it is doing exactly what it is told to do ("www.yelp.com/events"), but not what I would like. I like the idea of being able to do verbatim searches though! (and still miss the + button...)

It's encouraging to me that Google takes users' concerns into consideration. I prefer this dialogue to Google just 'getting things right in the first place' in fact. This is one example. Gmail's redesign is another. And I'm sure more changes are coming to both. I hope Google never stops making aggressive changes and listening to feedback aggressively.

Here's my Google Verbatim search engine plugin for your browser. http://mycroft.mozdev.org/search-engines.html?name=google+ve...

It might be just my home and work connections and their latency, but for me the auto-complete has hugely slowed down my searches.

I type rather fast (as you can tell by all the typos in my posts) and today I type something in and then wait for all the auto-complete options to settle down.

Sure, I don't have to hit the search button to get results anymore. But previously it was faster to type, then hit search, then get results. Much faster.

Mine returns faster than I can finish typing and look down.

I just added a custom Google Verbatim search engine to Chrome with keyword "gv", so I can type "gv testing" in the URL bar and get a verbatim search for "testing".

To do this, go to the "Manage Search Engines" section of Chrome preferences (Firefox has a similar feature somewhere) and use this URL:

https://www.google.com/search?q=%s&tbs=li:1

Woohoo! that's actually something that was annoying me greatly. Type stuff and Google just search for whatever it thinks is best. Bleh!

Anyway, that's awesome.

It's still stemming for me. I'm using my normal pathological search of [["everything wrong" crossfade]]. Verbatim mode roughly cuts the number of results in half but keeps giving me pages that only have the phrase [[everything's wrong]] and not the phrase [[everything wrong]].

I still love Google's search product but I am not real happy with the recent UI changes they have introduced. I particularly don't care for the preview arrow thing that now goes down the right side: I keep triggering the damn thing on accident.

Yes, THANK YOU! I find myself doing that as well. I don't understand why that is seen as an improvement.

Also, I always used the + to mean "the following word or quote surrounded string HAS to appear in the result text".

That seems different than quotes meaning "use this text exactly as i typed it".

I generally quote string searches but I do use pluses when I want to search for two separate strings in the same page, such as:

"Hacker News"+"Steve Jobs"

Also the cache link has been sadly moved in that panel.

If you'd like remove Preview then you can use the CSS from this post http://news.ycombinator.com/item?id=3194164 to disable it.

conradfr: You may also be interested in the CSS snippet in the above post since it will make the Cache link show next to the URL.

This is what I liked about the original method that Google used back in the 90s. It was always an exact text match for all words. At some point that changed, and to this day I still forget to use + to force the old behavior.

Well for a verbatim search of map(call())

I wouldn't expect google to filter out parenthesis. More than ever now that code search is dead.

"map(call())" would work either.

So much for verbatim ...

Wow 2/3rds of people use it wrong. That surprised me.

On the other hand, the post tries to pretend that this had nothing to do with +username or +brand and Google+.

No, it's the users, it's their fault, not our new product naming boo boo. We'll just silently break our flagship product and hope no-one notices.

Blame the customer seems a bad strategy.

Just wish they'd man up and admit it. Even if it's really not the reason, because that's the belief on the internet. It's times like this that remind me how Google really doesn't get customer service.

It's definitely the case that <0.5% of searches use the plus operator, and most of those searches aren't using plus correctly. If you used the plus operator correctly, you're well within the 1% of most savvy searchers.

You might be surprised and discouraged to discover that most people don't know about things like the minus operator to exclude terms from search results.

If you really want to be depressed, read this: http://www.theatlantic.com/technology/archive/2011/08/crazy-...

Could you explain how exactly the + operator is used incorrectly?

I know that personally i often end up with bizarre looking searches, purely because i know what i'm searching for. e.g. it might include purposeful spelling mistakes, requiring the plus to avoid autocorrection...but this doesn't mean i'm using the + incorrectly.

I just can't imagine somebody, potentially, understanding boolean logic, describing a search with it, and using it incorrectly. I can't even imagine the most naive usage being incorrect, unless you're sure that you know what they wanted to do somehow.

I'm at a conference and can't easily get on our internal VPN from here, but I'll try to remember to circle back around when I can get on our VPN.

But in general, you see a lot of stuff that looks like semi-random punctuation just sprinkled at the beginning of the query or throughout the query, things that look like people are using plus instead of space as if they're copying from an address bar, etc.

Ah ok that makes sense, thanks. I'm just mainly surprised at how little usage there was of the operator to begin with. I use it (used to use it) probably 20+ times a day, and would've assumed that it's common enough, at least in certain circles (academia etc).

Okay, I found the sample data. Lots of stuff like phone numbers: +XXX YYY ZZZZ. Lots of multiple plus searches, like +++++aaaa. Lots of people using + instead of space: aaa+bbb+ccc. Lots of searches with every word having a plus in front of each word: +aaa +bbb +ccc +ddd +eee. Some people surrounding words with plus: aaa +bbb+ ccc. Those were the first things that popped up in a quick glance.

Thanks a million man, so basically it seems like those people weren't actually meaning to use the operator at all. Sure you already parse them out anyway, though. Ah...i wont go on too much, i'm still surprised that so few people used it on purpose! Thanks again

To be sure, Google search used to allow the period to force a connection between terms. "foo bar baz" == foo.bar.baz

No wonder my website, foo.bar.biz has been getting so much traffic :o

I'd rather . notation than "" any day. Doesn't seem to work now...

Could it be that + is used by people who just don't know that Google treats it specially? Syntax of Google search is rather not common knowledge among general population. This might be a case especially when searching for books, songs, movies, magazines, etc. with '+' in title.

Example queries (random ideas):

  - 18+ movies
  - children + fire proverb
  - 2+3D magazine
  - C++
  - ++ operator

Personally I often just copy-paste general things I'm looking for (like song titles) from websites or IM conversations, with whatever characters they happen to be containing.

EDIT:

There are plenty of queries with + signs in autocompletion hints in Google search - like 18+ clubs, 2+2 forums, 0+ blood type, NaOH+H2SO4, etc.

At a complete guess, i'd guess that their parser would check for usage of a plus that doesn't fit the boolean mould and doesn't count it as such. And i guess too, because like you say...autocomplete knows about it, it has exceptions for acceptable +'s that also aren't included in the malformed searches that the Google guy is talking about.

So i still can't really imagine how there are so many misused ones...

OMG Finally. This will make one IT Guy's day job much easier, and the same guy's weekend/evening programming life easier too.;

too little too late?

It's hidden into the prefs when I'd liked it always on, on any device I use without having it tied to a specific account or anything.

I switched to ddg for regular and advanced queries, and use google for all the fuzzy stuff or very specific queries. It's still easier than to care about the verbatim setting.

What if you just want some words verbatim?

You'd just use +word right? Oh wait.

No, now you would use double quotes. So instead of searching for [hacker news ck2] you'd search for [hacker news "ck2"] to require the exact word "ck2". But remember, that could match anchortext pointing to the page. You can use "intext:" to require a match on the page.

Today's 0.5% is tomorrows 95.5%. It just takes time.

It is the so-called power users who lead the way for everyone else.

Doing away with a traditional boolean operator for the sake of marketing (of a copycat "social network" website no less) is not the sign of a company with a clear, intelligent vision.

But time will tell who is the wiser.

Actually, usage of the + operator has been in steady decline for years. With Altavista you had to use + to require a word. When I started at Google, many many more queries had tons of pluses.

Google's demographic was much more tech-savvy in the early years, but as we've struck partnerships with companies like AOL, the population of people searching on Google has become broader. I don't believe that 95.5% of users will be searching with + or double-quotes any time soon. But for the power users like people on HN, they now have an extra option that allows better slicing/dicing of search results.

You seem to be suggesting that Google's new demographic will never become more tech savvy or at least not "any time soon".

That's an interesting view.

I'm just a bit more optimistic.

I don't think there is any motivation for people at large to become more "tech-savvy". It's rather the case of machines becoming more human-savvy.. and I don't really see anything wrong with that.

Machines becoming more human-savvy is a slow and difficult process.

And people are generally impatient. They want things to work. Sooner rather than later. They want immediate results.

Whereas if someone has the interest, I can teach them a few tricks and make them incrementally more "tech-savvy" in a few minutes. They will see the results immediately.

The more "technical" life becomes, the more it stands to reason that more people will have an interest in becoming at least a little more "tech-savvy".

We know where Google stands on this.

Let's see what happens.

I definitely agree with you that it's probably faster to teach a single individual with a mild interest in becoming more techie than, for example, change an industry standard for search engines. But I'm talking about entire populations of humans and machines.

Changing an entire population of computer software/hardware is a far easier and faster process than changing the entire population of humans. We're much more stubborn creatures than we'd like to believe, plus we also have a longer lifespan thus making any dramatic cultural changes (techie or otherwise) a more or less generational thing. On the other hand, just look at how fast something like cell phones or the internet is changing on what now seems like a monthly basis.

People will always do the same things they've always enjoyed doing: eat at restaurants, socialize with friends, play games, listen to music... but in what way technology is involved with those activities will only be affected by how fast the technology can change to become more human-friendly, not the other way around.

Understood.

But here has Google has really made anything more human-friendly? They've simply removed a standard boolean operator in database search syntaxes because they noticed people were not using it (did people even know it existed?), or using it incorrectly. Does this make search more human-friendly somehow?

It certainly makes easy-to-type, using known standard operators, boolean searches more troublesome.

And the reason they have removed this operator?

Marketing. For something having nothing to do with search.

Oh I wasn't really commenting on Google's decision and whether it was right or wrong. Like you said before: "Time will tell". I was just offering an alternate view to your expressed optimism that Google's demographic (aka everybody) is becoming more tech savvy.

More like adds and then buries. I'd rather have the plus sign operator back, google.

Reply on news.ycombinator.com