I am designing a localized web app. I am leaning on auto-detect browser language setting. But I notice a number of respectable sites asking the user to select a language. Is there any usability issue you know of (from actual experiences out there) with just auto-detecting user language?
Thanks.
Give me a choice
Remember my choice
Use the auto-detect as default
Make transition easy
In many situation I prefer or even need the "original" over my local one, bad translations or different content being the major reason.
If you register multiple domains, you can base your auto-detect on that: When foo.com redirects me to foo.de, or otherwise shows me a german interface, it is actively ignoring my choice to go to foo.com.
MSDN did insist on showing me atrocious automatic translations and ALWAYS made me click to go to the readable, understandable english one (that's a step up: when they introduced it, the default selection for changing the language was something like Afrikaans).
Make transition easy: i.e. make it easy to go to the counterpart of the current page in a different language. Amazon often succeeds when I change ".com" to ".de", but then it fails to lead me to the german translation of the item. That's not always possible, as that requires each local view having the same structure and a 1:1 page mapping. But generally, you have to weight above requirements against other constraints of the project.
[edit] MSDN got better now :)
I would suggest to autodetect the language and display the site in this language or the default languge (probably english) if the translation is not available. Additionally present the user with a selection of languages on top or bottom of your page. The names of the languages should be written in the target language.
Don't do it like that: English, German, Italian.
But: English, Deutsch, Italiano.
Obviously there is the usability problem that you might detect a language that the user doesn't understand. How are you going to do the detection? Don't think everybody has their browser set to the correct language. IP-Adresses are also a very bad indicator for the users language.
Practical example: YouTube tried to convince me for a week or so to use the Japanese version, though I can't read Japanese. Not very helpful. Microsoft is also determined to serve me automatically translated versions of there documentation when I just want to read the English one.
So don't try to tell your users which language they're supposed to prefer, let them decide for themselves.
I really hate non-configurable auto-detection because a lot of applications are translated more than imperfectly. I would rather read perfect English than bad Russian. For example, some terms do not translate in a reasonable way, and trying to translate everything makes localized version faintly ridiculous.
Also some applications can not translate new features fast enough, leading to a mixed language.
So I always prefer to have a choice, and choose the version that is native to the application author -- for the best language (unless it is a language I do not know).
Update:
One situation when it has gone beyond ridiculous is DB2 (or its client tools, not sure), which forced me to install a Russian version, but all errors in this version were shown as "???????? ??? ??? ??".
Yes: at work, we have a Windows XP deployed with 'English' language (because we have worldwide site and only one kind Windows to deploy with only one kind of settings when it comes to language).
Yet all out applications must run in French. The auto-detect feature alone would not be enough for an appropriate display of the labels.
Sometimes when you are trying to describe something to a user over the phone and you are in a different location, it is very annoying when you are both looking at the same URL, but see different results. You might even go so far as to include the language in the URL similar to how wikipedia does it (e.g. en.wikipedia.org).
Also sometimes a user will be on a friend's computer and try to access a website but won't see it in their preferred language, because of the language settings on the computer.
I think the best solution would be to allow the user to override the setting, but default it to the auto-detected language.
I agree that the auto-detect is not enough.
Not many users know the settings for selecting their language. Therefore the settings will often be the default and therefore incorrect (for non-english users).
Related
We are migrating our app from iOS6 to iOS7 and we use programmatic way of creating view (rather from storyboard or nibs).
We are trying to support multiple countries with different languages.
Example,
English for - China, India, US
Simplied Chinese for- Taiwan, China
There can be custom override's for specfic country from the basic language localization set.
Now I need to have a common base for language bundles and country specific bundles.
Common Language Bundles: (base language bundles)
en.lproj
zh_hans.lproj
Country Specific Override Bundles: (if i have custom text for each specific countries)
ch(ina)_en.lproj
ch(ina)_hans.lproj
us_en.lproj
Problem:
Resource files (Translations) have to be duplicated for each countries(chinese, taiwan) with english, chinese. How can we avoid this ?. Images are also duplicated sometimes, it is a maintenance problem, if we start support more than 10 countries.
Android supports delta overrides of translations for each language translation per country, do we have anything in iOS similar to that ?.
I know it is not supported out of the box from iOS. What is the right way to achieve the same without duplicating the resources ?. Any hints or ideas to achieve the same ?.
Thanks,
Alex
I hope I've understood correctly.
1a.Image files will only need to be duplicated per language if they contain text or "imagery" that requires translation otherwise there should only be one version. From memory, you select which image files you want to be translated.
2a.A translation is needed for each language you want to support - there is no way round this (obviously). These usually live in "strings" files which you send off for translation.
2b.If you don't supply a specific translation for a string it defaults to the "base" translation. Unfortunately, I don't know how this would work with two "base" translations or even if this is possible as usually the base translation is the language you developed in. You will need to investigate further.
2c.You will need to manage deltas to your strings file yourself - through GIT perhaps? This is annoying but do-able although there may be third-party products that can do this.
In our company we are trying to decide whether to use english for our message id or whether we should invent some kind of key. We can only take the second approach if there are editors which allow us to work on multiple po files in one interface because translators will need to use english in order to translate into another language rather than translating from some cryptic message id.
Is there a tool which basically circumvents the messageid and allows translation from one language into another?
I guess many translation tools do support this mode as it is used by several translation formats (like Android resource strings or Java properties). However it is not commonly used with Gettext (as you might easily end up with application showing message ID instead of not translated text).
I believe it should work with Weblate, though most likely nobody has ever used this mode with Gettext files.
Update: This mode is fully supported in Weblate and used by many projects now.
I have a client who I am pricing an app for, however other than the English version they would also like a Japanese version. Has anyone had experience in a similar case, is there an easy way to do it? Do I need to create two versions, one English and one Japanese? If it were two Latin languages I could imagine it would be easier but Japanese write from top to bottom, right to left so this worries me.
You don't seem to know a lot about Japanese. They're perfectly accustomed to western-style left-right, top-down writing, especially due to the influence of computers. You can of course create separate views (views only, no need for separate apps) for Japanese that switch everything to top-down, right-left writing. But it's only a minority of apps that do that. In fact, the Daijirin Japanese-Japanese dictionary is the only example I know of.
Talk to your client what kind of Japanese localization he wants. Odds are, he just wants strings replaced. See #kelloti's answer.
As a general advise: Make sure you get a native translator/developer who can guide you in a good localization. Don't simply copy-paste in strings you get from somebody else that you have no idea how to even read. This only produces terribly localized versions.
Read the Apple documentation on internationalization. I don't think you should have many issues with Japanese (how else would they sell phones in Japan?)
I am developing applications for mobile phones with different operating systems (Android, Symbian, iPhone). Applications are sold internationally so they need to be translated to different languages in addition to english version.
I assume most mobile developers do the translations using some paid external service each time. This approach does not look very cost-effective to me. Would it make sense to have a website where simple translations would be done using crowdsourcing (other developers)? Most strings in mobile applications are very simple and short, for example "OK, "Cancel", "Are you sure?", "Please enter your password". Also the same strings are used in hundreds of applications. Instead of paying for translating all strings, developers could save money by only buying their difficult application specific translations.
Does anyone agree with this idea? I have seen many opensource projects doing the translations succesfully using volunteers.
I just found solution for me. Many users find this question in Google so I think my post must be helpful:
This is solution for us: crowdin.com - agile localization solution for tech companies
Microsoft allows you to view their terminology database: https://www.microsoft.com/Language/en-US/Default.aspx
That covers about 90 languages and will get you the things you mention such as common button captions, etc.
The problem you are facing after that is to try to get only the strings translated that you want. Most translators are going to charge you a minimum number of words. And they are going to want the entire resource file (regardless if you translated them yourself or not). Makes sense because localizing a product means that they need to have the whole picture to ensure consistency, etc. Professional translators will probably not charge you for what they call 100% matches.
I would never ever trust the translation of my product to crowd sourcing. Ever. You get what you pay for. Besides, just because you speak a language natively doesn't mean that you can write well, etc.
How do you check the crowd sourcing translation results for accuracy and quality? In a famous and documented occurrence recently the phrase "No lorries by this route please use the main road" was translated into "We are out of the office until Monday please contact us again then" and turned into road signs that were erected.
Crowd sourcing translation has been used and FaceBook is probably the largest company i know of that tried/used it. I have not tracked their progress but you could investigate it to see it's success or otherwise. Their method of quality checking was to get other people using the translations to vote for the one they preferred, so this was a case of crowd sourcing quality control. At this point the proposal that a camel is a horse designed by a committee jumps unbidden into my mind.
Translation, in spite of all the machine pumped into it, is still more of an art than a science. To translate correctly you need to have a native speaker translating from another language into their own. So for English to German you need a native German speaker who can speak English very well to do it. Within the profession very, very few translators will translate to a language in which they are non native. The reasons for this are many but boil down to the colloquial nature of language.
To be positive you could look at how Facebook fared and follow that route. Another route would be to approach not translators, but a translation agency, there are quite a number of these. Present them with the whole corpus you want translating in the original English and get them to quote you for the whole job. This would mean someone else manhging the job and the quality and they may have shortcuts, especially if the translations are to fairly standard "computerese" type phrases. i.e.'Home', 'Back', 'Next', 'Click here' etc.
I'm creating an English translation for a program written in German (i.e. all strings within tr("...") are German). Users who are in a non-English non-German locale will probably want to see the English translation, but with the program as it is now they will see German.
There are some ways to solve this problem:
Check if it's a German locale and force to English otherwise.
Present an option to the user.
Make the programmers change their source code to English.
What is considered best-practice for internationalizing where the source code is not in English?
These are two separate questions.
The best practice is to not use any kind of hard-coded string in the sources.
Strings should be stored in external files and loaded by ID.
But what you have there does not sound like the best practice. Might be too much work to get it there.
What you describe (the tr("...") stuff) sounds like gettext (or something similar).
That approach for gettext (and similar libraries) is that "the stuff in the sources is the ultimate fallback", used if the strings for the desired language are not present.
In this case I would go with "Present an option to the user."
You can't assume the user knows English.
Real example: in Switzerland the official languages are Italian, German, French and Romansh. If I ask for French and it is not present, then the next best option is probably German, not English. I Canada the official languages are French and English, so if I as for French and is not available, the next best option is probably English.
I think the best option is asking the user (during installation probably).
Change the source to English is too costly and not worth it. I live in Brazil, we have tons of codes in Portuguese and translating to English wan't necessary one time (we do make software to english speakers). Unless you have a client that requires you to do so (usually when you are selling the source also).
Hope it helps
OK, so I guess the three options are:
Recompile the program with translated strings.
This is fraught with danger as you'll end up with two copies of the source. Bug-fixes in one will need to be done in the other. And then, what happens if you need French? Italian? Spanish? The only advantage of this approach is that it's feasible for a non-developer to do the work. (Just about.)
Resource out the strings, and automatically check what the UI locale is on load.
Here the strings are replaced with GetResource("key") or similar. On load the program automatically translates to the user's culture. This might work, but I know plenty of German-speakers who have English-language culture installed on their PCs but who would prefer German language programs at some points.
Resource out the strings and give the user the choice on load
In general it's always best to give the user control. This might be a prompt on load, although if the application is used often this can be an annoyance. Perhaps a balance is to ask the user during installation for their preference and then give then an option in a dialog to later change this setting.
Note, by the way, that translation is not localisation. For instance: number formats are quite different in Germany (e.g. 1.233,44) from English (e.g. 1,233.44). Icons and suchlike often have national characteristics.