It’s time to address a longstanding issue with Google, and as these things often go, it has to do with Silicon Valley not knowing multilingual people are a thing.
A long, long time ago, searching for stuff on Google in different languages was a breeze. If you typed www.google.nl in your address bar, you went to Dutch Google. If you typed www.google.com, you went to English Google. If you typed www.google.de, you went to German Google. You may notice a pattern here – the country code determined your Google Search language. Crude, but effective.
Years ago, however, Google, ever on the lookout to make its users’ lives easier, determined, in its endless wisdom, that it would be a great idea to automatically determine your search language based on your location. Slightly more recently, Google seems to have started using not your location, but the information it has on you in your Google account to determine the language you wish to search in when you load Google Search, and on top of that, it tries to guess your search language based on the query you entered.
Regardless of whether I go to www.google.nl or to www.google.com, Google standardises to Dutch. The language menu in Tools is entirely useless, since it only gives me the option to search in “Every language” or “Only pages written in Dutch”. When I type in a longer, clearly English query, it will switch to showing English results for said query. However, with shorter queries, single-word queries, brands, or other terms that might transcend a specific language, Google simply doesn’t know what to do, and it becomes a game of Guess What Language This Query Is Parsed As.
As I’ve detailed before, Silicon Valley doesn’t get out much, so they don’t realise hundreds of millions of people around the world lead multilingual lives, speaking and searching in several different languages on a daily basis. Many Americans speak both Spanish and English on a daily basis, for instance, and dozens of millions of Europeans speak both their native language as well as English. Especially younger European generations have friends from all over the world, and it’s likely they converse in today’s lingua franca.
Of course, for me personally, the situation is even more dire. I am a translator, and especially when working on more complex translations, I need to alternate between English and Dutch searches several times a minute. I may need to check how often a term is used, what it means exactly, if a technical term is perhaps left untranslated in Dutch, and so on. I need to be able to explicitly tell Google which language to search in.
In its blind, unfettered devotion to machine learning and artificial intelligence, Google has made it pretty much impossible for me to use, you know, Google.
Meanwhile, DuckDuckGo has a really neat little switch right at the top of its search results, which I can click to switch between English and Dutch – I don’t even have to retype the query or reload the site from the address bar. The dropdown menu next to it gives me access to every single other language DuckDuckGo is available in. It’s difficult to overstate how this feature has turned web search from a deeply frustrating experience into the frictionless effort it’s always supposed to have been.
This tiny, simple, elegant little feature is what has drawn me towards using DuckDuckGo. I’m willing to accept slightly less accurate search results if it means I don’t have to fight with my search engine every single day to get it to search in the language I want it to.
I will continue to harp on Silicon Valley for barely even paying lip service to multilingual users, because it frustrates our entire user experience on a daily basis. To make matters worse, virtually all popular tech media consist of Americans who only speak English, assuring that this issue will never get the attention it needs.
I suspect this has to do with the Right to Be Forgotten mess. Google has identified you as a citizen of the EU so it is going to follow you where ever you go and make sure you don’t see those evil search results. As a side effect your language is being dragged along too.
No. This was from way before that.
And this right to be forgotten mess has an insane implementation. The EU has abdicated control of it to a panel of Google employees. Hint – Google is not the government, it should not be forced to make censorship decisions.
The correct solution to this was to take the offending website to a court of some kind (big or small or even a special one with email submission) and get a ruling. Then that court would order the hosting website to put a ‘do not index’ entry in the robots.txt for that site. Voil~A , now those pages disappear from every search engine.
It is just insane to have Google employees functioning as the court making rulings on this.
And don’t complain about pages hosted by companies outside of the EU. Gosh, if something is newsworthy enough to be hosted by servers around the world then it is not something that should be forgotten.
Edited 2017-11-26 23:27 UTC
I hate Google, and every service at A Valley, for this.
I live in a Caribbean Island and our language it’s Spanish but every person related with computers in the 80’s or 90’s in my country learned computer in. English, sometimes I don’t even know the translation in my language about something computer related.
So all my software and configuration are in English but I don’t speak English fluently so I would like to talk with my phone in Spanish, but as my configuration its in English I can’t talk with my phone in Spanish so I have to repeat every phrase in English like 3 times until Google understand it. I would prefer that Google allows me to configure several languages for several occasions.
Someday, someday I hope IA resolv this issue.
Just next to “tools”, there is a “settings” menu which will allow choosing one or multiple languages for Google search.
There is also advanced search, but it is even more hidden.
The “tools” menu seems to be a shortcut for some search options, but you’re right it is far from complete.
Doesn’t the No Country Redirect ( https://www.google.com/ncr ) feature work anymore?
That way you could at least use the Dutch and the US English searches next to each other.
boblowski,
I was not aware of this flag, but I did give it a shot after switching languages as described in my other post and it did not clear google’s region information.
Alfman,
I try to stay clear of anything Google, so I’m not sure how to properly test this, but from what I understand it should still work.
Just checked, /ncr does properly prevent redirection to the local Google search. But there are so many other factors influencing the search results, that it’s difficult for me to see what part was caused by the country/language settings in the first place.
The Google search interface is painful to use and has been for years. It’s the interface rather than privacy or search results which got me using DuckDuckGo years ago.
I recently had to use Google for a work-related issue and it was very awkward. I go to Google, click in the search box and start typing, the search box loses focus after a few letters. I click the search box again and keep typing, it loses focus. I click, it loses focus. I finally finish my query and see results, but cannot select them with my arrow keys, I need to use the mouse again. On DDG I don’t lose search field focus, there’s no annoying re-display of predictive search results and the site works great with a keyboard.
Some people claim DDG doesn’t have as accurate results, but I find the opposite. Maybe it’s because DDG doesn’t try to predict what I want and just shows me what I ask for. Google usually gives me a bunch of unrelated topics based on my region rather than what I want to see.
It’s true that sometimes Google is better than DuckDuckGo. Although DDG is my default search engine, sometimes Google is better. In these occasions typing ‘g! ‘ before your query in DDG is a great help. It sends your query to the encrypted version of Google search.
jessesmith,
Out of curiosity, what are ‘accurate results’? As far as I know every modern search engine just gives ‘suggestions’.
Is there any search engine nowadays that returns actual verifiable results?
I guess it all depends on what you search for.
For me, DDG DOES give accurate and verified results. I work on Linux a lot, and it gives code bit breakouts with the most voted answer from stackexchange when you look up computer questions.
Granted there are the few times when those are wrong/old answers, but it’s more the fault of contributors to stackexchange over DDG.
leech,
I see these results in DDG too.
Unfortunately I feel the stackexchange site itself has suffered and stagnated over the years. Very frequently I find exactly the question I’m looking for only to find out that SE has closed it prematurely for bogus reasons. I don’t know if this is caused by a poor incentive system that rewards excessive & unhelpful action and even group think, but it seems like SE mods tend to assert their control without really considering how useful the questions & answers are.
Sometimes I feel like I could do an excellent job answering some of the questions people have. However I’m actually very glad I’m not a part of the SE community because while I really do like helping people, I totally despise heavily bureaucratic processes. SE is full of it and I know it would make me miserable!
Edited 2017-11-27 22:21 UTC
Sure, DDG and others do a remarkable job guessing what you want based on your input, but that was not what I was thinking about.
Sometimes I’m interested in the search itself, not just in the top result. To stick with Thom’s translation business, perhaps you want to know how often a certain exact spelling is used or if a technical term is related to a specific domain or language.
There was a time when you could run reasonably exact queries in search engines, even in Google. And the search results also were repeatable. We could enter the same query and get similar results.
I was just wondering if there is any search engine out there now that offers any functionality like that.
This is all you somehow. It doesn’t do that for anyone else. I’ve seen some badly written browser extensions screw with focus like that and sometimes Windows accessibility programs which seem to be trying to “help”. Maybe some sort of on-screen keyboard or other input “helper.”
It could also be a bad Linux window manager. I’ve seen XFCE and Enlightenment do very, very broken things in the past although I don’t know what state they’re in these days. And of course if you run things like dwm or awesome or i3 you are entirely on your own to guess what will happen.
Thom,
I’m not sure if this will help or not since I speak neither Dutch nor German, but it’s worth a shot. Startpage.com (also known as ixquick.com) has a Language setting in the menu, and it uses Google as its backend. I use Startpage to escape the Google search bubble but it may be useful to you as a translator, giving you Google’s results with more control.
Also, you can use the !g command to search on Google with DDG, though I have no idea if your language settings will make the jump.
I was experimenting access google using multiple country TLDs (from the US). Google ignores the TLD and continues to serve up the english version of the website.
I noticed that after clearing the cookies (and only after clearing the cookies) and navigating to google.fr / google.nl / etc, I had the following option “Google offered in: Fran~A§ais / Nederlands /etc”. Clicking it opened up access to a french home page with french results that did not originally show up on google’s english site even using the exact same queries! So switching google to the correct language is important to do before querying – google either does not do this automatically, or it is just broken.
Once the language cookie gets set, google very stupidly removes the link to change the language again even when switching TLD. This is either a bug or just poor design.
I found that the language switching links that are displayed when cookies are cleared do continue to work even though google isn’t displaying the links to switch language. So Thom, it may be worth trying to bookmark them and using the bookmarks to switch languages as needed that way.
Edited 2017-11-27 02:18 UTC
If it’s a bug than it has been there for many, many years.
A much much worse phenomenon exists on the internet and that is companies completely ignoring your web browser language settings. This madness began approximately 15 years ago and now I don’t remember which company decided to be the first, however all others followed.
As odd as it sounds, I can actually kind of understand this move, for two specific reasons:
1. )It’s generally non-trivial to change your preferred language reported by your web browser in request headers (unless you grab a header injection extension). On Windows, you have to change the system display language. On Linux and BSD (and I would assume OS X), you have to start the browser with a particular value in the ‘LANG’ environment variable (and usually can’t easily run separate instances with different values).
2. You aren’t necessarily going to always be using your own device to access your account on any given site, and most people object to having the display language changed on their system (or you just can’t, if it’s a public terminal), so it makes sense to have the site itself store the language settings for the account so that they follow you.
That said, sites should ideally default to whatever the first language they support in the language header list in the request is, but that’s not exactly the same thing.
Not neccessarily. Browsers like Opera (here: on FreeBSD) can handle request header generation independently from UI settings (controlled by $LANG, $LC_ALL or the more specific $LC_* settings). You can even specify multiple languages as a preference list (for example, german first, english second, and the web page would provide a german interface if it has, and if it hasn’t, use an english one).
I find it outright stupid to personally register (or even identify) myself to web services as basic as a web search engine just to tell it what language I want to use to search. Using the the information from the TLD (for example .de or .com) is an easy approach, but it won’t always work, especially for countries that have more than one official language (like Switzerland: de, it, fr). In such cases, simply specifying a parameter via URL would be an easy solution, like issuing google.com/?lang=fr from Switzerland by a user who wants to perform a search in the french language; googe.com without any parameter could simply default to english results, because… why not?
Is this a too much simplified approach? I have no idea…
The problem with this is that most sites actually have differing content based on region (and by extension, language). For handling multilingual countries, it would work, though I don’t know of many other than Switzerland (with four national languages, German, French, Italian, and Romanish (listed in order of usage)) where you have significant populations of monoglots in a country with multiple national languages. For any site engaging in some form of commerce though, it becomes very important to display information pertinent to the country the user is intending to purchase things from.
On Linux Firefox it’s in Preferences, General, Language and you create a prioritized list of languages. Yes, that doesn’t work with running different settings at the same time. You’d need to run separate profiles for that.
I think this was made a while ago by countries’ ISPs and governments controlling traffic to and from servers. Before this had happened, the internet was really open: you could access any site anywhere.
But now there are firewalls in play and traffic shaping to control what information gets disseminated to whichever country.
Private companies can control this on their servers as well to reduce DDoS attacks from other countries, or perhaps as a cheap solution to selling to people from other countries stuff that can’t be sold for one reason or another in another country (licensing, local laws).
Little more worried about fixing my water heater and my furnace before I freeze to death… after that i’ll happily help you with your multilingual issue….
Having a need to search in 2 Languages I tend to use two different browsers (FF&Chrome) with different Google accounts.
Additional benefit are search results as I search for in native language for one area of interests (e.g. hobby) and in English for another area (e.g. work).
Yesterday I used the “find in page” (Ctrl+F) functionality in Google Chrome. Imagine my surprise when “~A~”, which I typed in, matched “O”! Those are two separate letters! It’s like having E match F, because E is an F with an extra line in it.
Moreover, T and L should match I, and R should match P, for the same reason.
And d should match b, because it’s just a b backwards.
I just tried it in my browser (Vivaldi), and you’re not lying – ~A~ also highlights O! I quickly tried Word, and Word does NOT highlight O when searching for ~A~.
This is a whole new can of worms and probably an upcoming article so thank you.
…wtf even.
I don’t know much about Dutch specifically, but I was under the impression that it was not hugely unusual when dealing with ASCII-only services in languages with non ASCII letters to just omit diacritics or convert to a ‘romanized’ letter, and have people infer from context the intended meaning. Such usage is at least not hugely unusual in romance languages and German for specific cases, though it usually results in an alternative orthography (for example ‘ss’ for a sharp ‘s’ in German, or using ‘x’ after a letter to indicate a circumflex in Esperanto (because x isn’t used otherwise in the alphabet)). If so, this is probably ultimately a holdover from that (and likely also counts on the fact that most people search for whole words).
Those are not “diacritics”, they are separate letters with their own sounds and positions in the alphabet.
In a context where they are not available, they are translated like this:
~A| -> ae
~A, -> oe
~Ayen -> aa
This would have been reasonable, as there aren’t many places that you can find these combinations inadvertently.
However, Chrome interpretes them as:
~A| -> ae
~A, -> o
~Ayen -> a
It’s not even consistent.
Bug report about diacritics: https://bugs.chromium.org/p/chromium/issues/detail?id=71741#c37
That is exactly consistent with how most native English speakers are likely to type each of them, and I would guess how most people who don’t have the given letter in their language are liable to do so as well (provided of course that they use the Roman alphabet). From a practical perspective, that’s actually exactly how I would expect them to be interpreted when I’m trying to quickly search some text for something, provided there is nothing indicating otherwise in most cases (the exception being where alternative orthographies are so common that they’re becoming standard in all cases, such as the phasing out of ‘~A"Y’ in favor of ‘ss’ in modern German).
Just tried it in Safari and ~A~ finds only ~A~. Not that Safari doesn’t have its own set of issues and if you are on Windows or Linux it doesn’t really matter.
Well, I can’t vouch for the sanity of this, but this is a thing that is often used to figure these things out:
https://docs.oracle.com/cd/E29584_01/webhelp/mdex_basicDev/src/rbdv_…
Python and perl have librries for doing this, they all map ‘~A,’ to ‘o’
If this is not correct, as you say it is wrong, I’d suggust putting in a pull request for
http://search.cpan.org/~sburke/Text-Unidecode-1.30/lib/Text/Unideco…
https://pypi.python.org/pypi/Unidecode
As a native Danish speaker, working in English and living in Germany. I have to give up on many websites, many pages including most of googles are routinely completely broken. And trying to set a language or a county just makes things worse, most sites just refuse to handle the combination.
Even Windows speaks a wierd pidgy combination of Danish, English and German. Just because my keyboard is Danish, my configured language is English and my locale in Germany.
I am an American currently living in Bulgaria.
When I use google.com, I get my results in Greek.
I’ve crossed the border over to Greece a few times. I think that confused Google and it’s been messed up since.
It’s been happening for so long that it’s no longer funny.
Besides being a pain with general functionality it can also allow hackers to spoof urls because of how some browsers handle unicode characters (i.e the incident with apple.come vs xn--80ak6aa92e.com ).
It’s been fixed in recent builds of Chrome but who knows what else can come up from this..
Edited 2017-11-27 16:11 UTC
Sidux,
https://xn--80ak6aa92e.com
Wow, that’s amazing, they’ve got an https certificate and everything. In firefox it looks exactly like “https://D°~NEUR~NEUR'ODu.com/“ and to be perfectly honest that would have fooled me.
Edited 2017-11-27 16:36 UTC
True
Https does not guarantee the authenticity of a website by any means. That’s another fake information promoted everywhere on mass media.
Https only guarantees that the traffic between client and server is private and cannot be spoofed easily.
As someone who travels a lot, i find choosing language based on location is monumentally stupid…
Especially since web browsers have for years been able to send an accept-language header… Why then does google totally ignore the languages i have explicitly stated i want, and chosen to serve me something i can’t understand purely based on the source ip address i connected from?
Google’s entire premise for their UI the past few years is that people are fucking retards who can barely hold a mouse or use a touchscreen correctly, and so you should try to guess what they actually want rather than what they directly requests.
I just notice Google doesn’t always honor the + modifier to require words in search results.
WTF is the point of it, then? Why is it still documented as something that works?