TOP

What Good is Google Translate?

Whenever I tell people what I do for a living, the first question I get asked is: “So how many languages do you speak?” (A: 2)  The second is often what my opinion of Google Translate is. If I’m concerned that it will make our industry obsolete (not really), what the quality of translation is (varies), and isn’t it amazing? (Yes) Google Translate works by analyzing web pages and other online content that is already translated into multiple languages, and using this content as its “translation memory”.  So in reality, its output is only as good as the input… and being that the internet as a whole is its input… you get the idea.  Individual users can suggest better translations, which seem to have helped its quality, but you still get pretty quizzical translations sometimes.  It’s also important to remember that anything you feed into GT might become the property of Google forever, so if you have any confidential or sensitive information, you at least want to take out any identifying company names, or names of individuals out of any text, if you decide to use it at all.

Having said all that, GT is an amazing tool. I use it at least a few times a week. The main ways I find it useful is for the purposes of:

1.  Verifying Text Placement – Translators tend to regard themselves as artisans, more than technologists. To this day, when working with certain translators, they ask if they have to use a computer-assisted program rather than just translating “as they see it” (yes, yes you do).  As a consequence, even though our translations are seen by the initial translator, their proofreader, and another translator, we still have to take the additional step as project managers to make sure that nothing got missed (especially white text), that there are as many paragraphs/numbers/bullets as the original, and that the text all fits where it’s supposed to (doesn’t get cut off due to text growth).  I use GT to identify target lines of text against the source, to make sure the text is in the right place.

2.   This just looks wrong! I had a Kannada translation the other day, and part of it was a date. However, the portion that was supposed to be the date seemed to be in the wrong cell.  I ran it through GT, and found that the date was indeed in the wrong cell. As it turned out, the translator informed me that because word order is different in Kannada than English, they did it that way so the text would display correctly. But I’m still glad I checked.

3.   What does this say? vs. what does this mean? GT provides very “Gist”-y translations.  It’s useful if you find an article, or document in another language, and want a direct translation just to get a general idea of what it’s about.  However, professional translators aren’t looking for verbatim word-for-word equivalencies in the target language, they’re looking for a way to effectively communicate the meaning to a reader who speaks the target language (and almost always belongs to another culture).  For example, to say “I dropped the ball” in Spanish, you wouldn’t say “yo lo cae la pelota” (literal) you would say instead: “Se me cayó la pelota”, or the ball got dropped.  GT has been improved enough where if you enter in English “I dropped it”, you get the correct Spanish translation (color me impressed!), BUT if you enter the incorrect “yo lo cae la pelota” (SPA>ENG), guess what English translation it gives you?  Similarly, if you enter “I missed my plane,” it gives you “Me perdí mi avión” (it should be “vuelo”… “avión” makes sense, but almost no one will ever actually say this. I found 3 hits over all of Google when I entered it as a search term, vs. 383,000 for the correct translation).  Now, if you enter “I missed my flight”, it gives you the correct translation. But you have to be expecting these issues in order to get the correct output.  A non-Spanish speaker is going to get some pretty inaccurate results, or at least funny-sounding ones, eventually.

GT is an amazing tool. I can’t stress that enough.  Some in the translation industry have been offering proofread Google translations as a way to cut customer costs, in an industry that often finds it difficult to explain its overhead to its customers.  We find that this approach restricts the translator too much, and it also makes it difficult to work with our existing translation memories for our long-time clients, so up until now GT post-editing hasn’t been a service that we offer.  We have to advise against ever using GT by itself for anything that’s going to be disseminated, whether on a public sign, or even an internal document without the caveat: “This came from Google Translate”.  I’ll leave you with one more funny Spanish-English example. I entered “I am full” into GT, and I got “Estoy lleno”.  If I’m a woman, it’s “estoy llena”… but in many places that means “I’m pregnant”!  Use Google Translate as the amazing tool it is, but recognize its limitations.  Or be prepared for some embarrassment.

Read More
TOP

Translation Memories, MT and the Future of Language Services (Part 2 of 2)

“Translation” ; “Cross-Cultural” ; “Cross-Cultural Training” ; “Global Diversity” ; “Diversity” ; “Intercultural” ; “Intercultural Competence” ; “Localization” ; “Global” ; “Global Ethics”; “Globalization” ; “Global Business” ; “International Business” ; “Contextualization” ; “Cultural Adaptation” ; “Cultural Audit” ; “Global Delivery”; “Internationalization” ; “Inclusion” ; “Intercultural Training” ; “Global Cultures” ; “Global Business Skills”; “Global Workforce”; “American Culture” ; "Machine Translation"

Source: BoingBoing.net

One of the biggest issues for translators and LSPs, is the lack of knowledge their buyers may have, or rather, the translation industry’s failure to communicate their knowledge to their buyers in a way that resonates with them.  Part of the problem is that learning a second language is often difficult, time-consuming, and not always effective without true immersion in a language & culture.  It’s particular problematic when working with Americans or American companies, as only 9% are fluent in a language other than English (nearly 50% of Europeans speak two languages).  (Sources: US Census 2000, European Commission Directorate General for Education and Culture, respectively)

If a person has never lived abroad, learned a second language, or experienced being a cultural/linguistic minority, differences between languages and cultures can be totally lost on them.  They may adopt a Universalist perspective: “we’re all basically the same,” and tend to have an attitude minimizing the differences between themselves and those from different cultures.  The best-educated of translation purchasers, as educated about the process as they may be, often don’t speak another language, and the challenges of working across languages will often not be intuitive to them.  

(more…)

Read More
TOP

Translation Memories, MT, and the Future of Language Services (Part 1 of 2)

“Translation” ; “Cross-Cultural” ; “Cross-Cultural Training” ; “Global Diversity” ; “Diversity” ; “Intercultural” ; “Intercultural Competence” ; “Localization” ; “Global” ; “Global Ethics”; “Globalization” ; “Global Business” ; “International Business” ; “Contextualization” ; “Cultural Adaptation” ; “Cultural Audit” ; “Global Delivery”; “Internationalization” ; “Inclusion” ; “Intercultural Training” ; “Global Cultures” ; “Global Business Skills”; “Global Workforce”; “American Culture” ; "Machine Translation"There’s been quite a bit of discussion lately about machine translation and the ultimate role of the human translator in the rapidly changing translation and localization industry.  Machine translation tools (like BabelFish and GoogleTranslate) are extremely fast, cost effective, and usually open-sourced.   However, along with the benefits come disadvantages as well.  First, the text isn’t encrypted, so if you have privacy concerns, using either tool is not advisable.  Second, particularly in the case of Google, they won’t release the contents of their translation memory, and any text that you casually run through the translation engine, and any edits or final products produced by translators using the engine become Google’s property… forever.  Essentially, you’re giving them content for free.  There’s also a danger in crowdsourcing, as there  have been cases of renegade “translators” inserting their own “translations” into crowdsourced platforms like Facebook, resulting in profane language on customer interfaces.   

All issues aside, it’s pretty clear that some form of MT is going to play a major role in the future of translation.  While machine-generated translations are always going to require a final human review due to the fluid nature of language and the nuances of context and culture, the cost-savings that can be realized by using MT are significant.  Even at a rate of 25 cents per word for high-level translating and proofreading, a 5,000 word document will cost around $1,250 per language.  Now, take the same content and run it through an open source translator, and have a skilled translator post-edit what the MT produces… even if this content is complex and takes the translator a whole day, the cost savings can still be significant.  

(more…)

Read More