Re: [MCN-L] Multilingual websites
Hi All, This is Tim Schwartz from https://alley.co – thought I would chime in on how our agency typically works through this process. We have built many multilingual websites over the years, typically with WordPress. Our go-to method is to use Google Analytics to look at the percentages of various languages are visiting the site. This is the first big aha! From there we decide on three categories: 1. Languages where the navigation and visitor information will have human-entered content for each language (this might be your top 3 to 4 languages as shown in Google Analytics, but this could be larger, depending on the tourist visitors—a good ol’ in-person visitor survey can help you determine these) 2. Languages where it is really important that any type of post content can be entered in that language (think of a true bilingual site that offers English and Spanish versions for every post) 3. Languages where we decide not to build any special support for this language (sorry visitors) We then usually build language options into the taxonomies, pages, and custom post types themselves, so that each title or term can have various language versions (this is how we handled https://agsiw.org/ and https://www.brookings.edu/ ). Currently, WordPress VIP is recommending the plugin MultilingualPress - https://wordpress.org/plugins/multilingual-press/ as a good tool for managing multilingual content. Whether you go with a custom language implementation or use a plugin off the racks, it is very important that any code that renders words be passed through a translation function. This also makes it so that as new content is added, the navigation changes, your visitors change, you can easily add new languages and translations. We do not recommend building machine generated translations into a site. Browsers and browser plugins (especially Google Chrome) are capable of doing the machine translation for you. This technology is becoming increasingly integrated into browsers by default. For example, a browser knows your preferred language and the language of the content on the site you are visiting, so it can offer the machine translation for you. This makes it no longer necessary to include machine generated language options for site visitors. Further more, as everyone has pointed out, machine translations can be horrid, there is no reason to bake these into the site, rather let the browser do it, but only if the user asks for it. Don’t forget, someone might have their browser set to Spanish but also speak English. Should you automatically serve them machine translated content? I think not :) Figuring out how to make sure that your site and visitor information can be understood by anyone who comes to your site is an important problem, but one that is better solved by building a real solution for those visitors that make up a higher percentage of website traffic and let the visitors browser or browser plugins solve for the smaller percentages. See you all at MCN! Cheers, Tim > On Oct 23, 2018, at 6:03 PM, Mark Andrews wrote: > > This is all super helpful. I love how willing you all are to share your > expertise. It strikes me that this is an important way for institutions > like ours -- that serve so many people for whom English isn't their first > language -- to welcome everyone and increase access, and that I'm woefully > under-knowledgeable about the topic! > > Thank you! - Mark > > On Tue, Oct 23, 2018 at 1:18 AM Amalyah Keshet > wrote: > >> I remember reading two or three years ago that Google researchers had >> developed an AI component for Google Translate, which revolutionized it. >> People woke up the day after the (secret) release and reported "hey, what >> happened? this is infinitely better!" I gave it a try at after reading >> that, and sure enough, there was a remarkable improvement -- although some >> lamented the demise of a popular source of amusement. I use it for >> Hebrew-English translation -- not a small challenge. >> >> No one assumes that Google Translate will do a masterful job. It's a tool, >> not a professional translator and editor. (Just like Google is a search >> engine, not a professional research assistant). Perfect for landing pages >> on things like museum websites, "contact us", etc. Less so for exhibition >> texts, articles, and deep content, but its certainly better than nothing. >> If you have an fat budget for multiple-language translation and editing >> staff or services, fine. Most don't. (BTW it's a great time-saver for >> creating a draft that a human can then edit and improve.) >> >> Deliberately not putting the widget on a website seems the opposite of >> customer service. Yes, the visitor can copy and paste into Google >> Translate him/herself, but just having the widget there says "we want you >> to stay and read and understand us. Maybe this will help." >> >> Amalyah Keshet >> >> >> >> >> >> *Amalyah Keshet*
Re: [MCN-L] Multilingual websites
This is all super helpful. I love how willing you all are to share your expertise. It strikes me that this is an important way for institutions like ours -- that serve so many people for whom English isn't their first language -- to welcome everyone and increase access, and that I'm woefully under-knowledgeable about the topic! Thank you! - Mark On Tue, Oct 23, 2018 at 1:18 AM Amalyah Keshet wrote: > I remember reading two or three years ago that Google researchers had > developed an AI component for Google Translate, which revolutionized it. > People woke up the day after the (secret) release and reported "hey, what > happened? this is infinitely better!" I gave it a try at after reading > that, and sure enough, there was a remarkable improvement -- although some > lamented the demise of a popular source of amusement. I use it for > Hebrew-English translation -- not a small challenge. > > No one assumes that Google Translate will do a masterful job. It's a tool, > not a professional translator and editor. (Just like Google is a search > engine, not a professional research assistant). Perfect for landing pages > on things like museum websites, "contact us", etc. Less so for exhibition > texts, articles, and deep content, but its certainly better than nothing. > If you have an fat budget for multiple-language translation and editing > staff or services, fine. Most don't. (BTW it's a great time-saver for > creating a draft that a human can then edit and improve.) > > Deliberately not putting the widget on a website seems the opposite of > customer service. Yes, the visitor can copy and paste into Google > Translate him/herself, but just having the widget there says "we want you > to stay and read and understand us. Maybe this will help." > > Amalyah Keshet > > > > > > *Amalyah Keshet* > *Copyright Management for Cultural Heritage, Jerusalem, Israel* > *Senior Consultant, Naomi Korn Associates, London * > > > > > > > > > On Tue, Oct 23, 2018 at 4:35 AM Matt Morgan > wrote: > > > We had the Google Translate widget on the Met's site, in just the visit > > section, for a while during my time there, and then we added it to a > > part of NYPL's site but I never managed to get it added to the whole > > site and actually lost my job there because I wouldn't shut up about it. > > True story! (I've had LOTS of other jobs since then, don't worry.) > > > > Anyway, the problem with human translation is this: you're already > > pretty careful to get the English right, and probably a few people are > > involved in the production and editing of any given page. Some or all of > > them may actually be trained & experienced as professional editors in > > English, and the educators/curators/content owners probably speak and > > write English pretty fluently, perhaps even well. Basically, you spend a > > lot of time getting the words right. How are you going to meet that > > standard in even one other language? You're going to hire some outside > > person who may or may not be good--it's hard to tell because none of you > > are professionals in that language--and not edit it nearly as carefully > > as the English because who's qualified to do that? And then it's so much > > extra work, when you change the English a little bit, you'll probably > > either just change the other languages in-house, with whatever local > > speakers of that language you can find, or maybe put it off for now > > since it really wasn't that big a change anyway. Over time you'll know > > that the translations are kind of falling behind the primary pages, but > > it'll be OK because everyone else is paying less attention to them than > > you are, so it'll just become this thing that eats at you a little bit, > > but you take comfort knowing that at least it's better than having no > > translations at all. > > > > In case it's not obvious, if you're Canadian substitute "French and > > English" for "English," and if you're Belgian or Moroccan or whatever, > > substitute the four or six languages you're a pro at. However many it > > is, it's not 100 and it's probably not even eight or nine. > > > > The problem with machine translation is exactly what Susan points out. > > It has issues. Even when it's pretty good it's not very good at > > technical terms, or colloquialisms, or proper names, or a dozen other > > kinds of repeated issues. And your art or history or science museum > > website is going to be about 30% those things. Probably 50% at the > > Exploratorium. > > > > Here's the thing: even with all those issues, it still helps. If you > > think of machine translation just as an aid to comprehension, and you > > think of adding the widget to your site as a convenience, just to make > > it easier for visitors to do something they can already do and you can't > > stop them anyway, why not? Well I mean, that's what I thought and I lost > > my job. But I still think it's a logical argument. And I haven't seen a > > whole lot of museum
Re: [MCN-L] Multilingual websites (Amalyah Keshet)
> > Less so for exhibition texts, articles, and deep content, but its > certainly better than nothing. Just a word of caution here for interpretative text. IMHO nothing is better than machine translation if dealing with culturally significant text. Do you really want to trust a machine to translate nuanced interpretation about ethnic cleansing or colonisation? This may land your organisation in some sticky situations if the translations somehow trivialised certain historical events. Cheers Glen ___ You are currently subscribed to mcn-l, the listserv of the Museum Computer Network (http://www.mcn.edu) To post to this list, send messages to: mcn-l@mcn.edu To unsubscribe or change mcn-l delivery options visit: http://mcn.edu/mailman/listinfo/mcn-l The MCN-L archives can be found at: http://www.mail-archive.com/mcn-l@mcn.edu/
Re: [MCN-L] Multilingual websites
I remember reading two or three years ago that Google researchers had developed an AI component for Google Translate, which revolutionized it. People woke up the day after the (secret) release and reported "hey, what happened? this is infinitely better!" I gave it a try at after reading that, and sure enough, there was a remarkable improvement -- although some lamented the demise of a popular source of amusement. I use it for Hebrew-English translation -- not a small challenge. No one assumes that Google Translate will do a masterful job. It's a tool, not a professional translator and editor. (Just like Google is a search engine, not a professional research assistant). Perfect for landing pages on things like museum websites, "contact us", etc. Less so for exhibition texts, articles, and deep content, but its certainly better than nothing. If you have an fat budget for multiple-language translation and editing staff or services, fine. Most don't. (BTW it's a great time-saver for creating a draft that a human can then edit and improve.) Deliberately not putting the widget on a website seems the opposite of customer service. Yes, the visitor can copy and paste into Google Translate him/herself, but just having the widget there says "we want you to stay and read and understand us. Maybe this will help." Amalyah Keshet *Amalyah Keshet* *Copyright Management for Cultural Heritage, Jerusalem, Israel* *Senior Consultant, Naomi Korn Associates, London * On Tue, Oct 23, 2018 at 4:35 AM Matt Morgan wrote: > We had the Google Translate widget on the Met's site, in just the visit > section, for a while during my time there, and then we added it to a > part of NYPL's site but I never managed to get it added to the whole > site and actually lost my job there because I wouldn't shut up about it. > True story! (I've had LOTS of other jobs since then, don't worry.) > > Anyway, the problem with human translation is this: you're already > pretty careful to get the English right, and probably a few people are > involved in the production and editing of any given page. Some or all of > them may actually be trained & experienced as professional editors in > English, and the educators/curators/content owners probably speak and > write English pretty fluently, perhaps even well. Basically, you spend a > lot of time getting the words right. How are you going to meet that > standard in even one other language? You're going to hire some outside > person who may or may not be good--it's hard to tell because none of you > are professionals in that language--and not edit it nearly as carefully > as the English because who's qualified to do that? And then it's so much > extra work, when you change the English a little bit, you'll probably > either just change the other languages in-house, with whatever local > speakers of that language you can find, or maybe put it off for now > since it really wasn't that big a change anyway. Over time you'll know > that the translations are kind of falling behind the primary pages, but > it'll be OK because everyone else is paying less attention to them than > you are, so it'll just become this thing that eats at you a little bit, > but you take comfort knowing that at least it's better than having no > translations at all. > > In case it's not obvious, if you're Canadian substitute "French and > English" for "English," and if you're Belgian or Moroccan or whatever, > substitute the four or six languages you're a pro at. However many it > is, it's not 100 and it's probably not even eight or nine. > > The problem with machine translation is exactly what Susan points out. > It has issues. Even when it's pretty good it's not very good at > technical terms, or colloquialisms, or proper names, or a dozen other > kinds of repeated issues. And your art or history or science museum > website is going to be about 30% those things. Probably 50% at the > Exploratorium. > > Here's the thing: even with all those issues, it still helps. If you > think of machine translation just as an aid to comprehension, and you > think of adding the widget to your site as a convenience, just to make > it easier for visitors to do something they can already do and you can't > stop them anyway, why not? Well I mean, that's what I thought and I lost > my job. But I still think it's a logical argument. And I haven't seen a > whole lot of museum websites succeed at human translation in more than a > couple languages, except on very limited sets of pages. > > There have been a couple major museum websites that provided the Google > translation widget on every page, as a convenience. The two that I > recall having it in the past no longer seem to provide it, so I have to > imagine they similarly had a champion for it who moved on (willingly, I > hope!) and their successors were less interested in fighting that fight. > > Anyway, here's my suggestion: try Google on a few of your pages in any > languages for
