Goodbye, Language Barrier?
Our reader tested new translation software. Have you tried these or other language-translation software? What did you think?
From Ken Clark "Goodbye, Language Barrier" nicely illustrates the problems business
people encounter when they first attempt to use machine translation. But
Nick' is really mistranslating if he blames his problems on the software.
A machine translation tool out of the box is of course no way to write a
press release. Skilled and expensive translators are required for that. So
why does machine translation account for the vast majority of translation
done today? Because it's fast and cheap, and even in it's raw form often
better than no translation at all
"Quick and dirty" machine translation solutions are most effective when
used:
1. For a quick-fast read of inbound communication.
2. For review of foreign language material to find relevant information in
order to reduce very large data sets.
3. As part of a workflow to reduce the cost of human translation.
Quality also varies widely by language. The options available in the more
common commercial languages such as French, German, Spanish and Japanese are
far superior to the tools available even for such widely spoken languages as
Chinese and Arabic.
To get really high-quality machine translation, careful attention to input
and output is required. Controlling source content, careful construction and
maintenance of domain-specific glossaries and validation by human editors
can make a dramatic difference in quality and potential applications. This
enables the creation of machine translations indistinguishable from
human-only efforts at a fraction of the cost. Those are the machine
translations that no one complains about.
Granted, you can't get that out of the box. It takes a team with souped-up
versions of these translation tools and the expertise and knowledge to pull
it off. But small businesses that want to grow big should not write off
machine translation just yet.
From Ken Clark, New York, NY "Goodbye, Language Barrier" nicely illustrates the problems business people encounter when they first attempt to use machine translation. But Nick' is really mistranslating if he blames his problems on the software.
A machine translation tool out of the box is of course no way to write a press release. Skilled and expensive translators are required for that. So why does machine translation account for the vast majority of translation done today? Because it's fast and cheap, and even in it's raw form often better than no translation at all.
"Quick and dirty" machine translation solutions are most effective when used:
1. For a quick read of inbound communication (great for customer chat and e-mail).
2. For review of foreign language material to find relevant information in order to reduce very large data sets.
3. As part of a work flow to reduce the cost of human translation.
Quality also varies widely by language. The options available in the more common commercial languages such as French, German, Spanish and Japanese are far superior to the tools available even for such widely spoken languages as Chinese and Arabic.
To get really high-quality machine translation, careful attention to input and output is required. Controlling source content, careful construction and maintenance of domain-specific glossaries and validation by human editors can make a dramatic difference in quality and potential applications. This enables the creation of machine translations indistinguishable from human-only efforts at a fraction of the cost. Those are the machine translations that no one complains about.
Granted, you can't get that out of the box. It takes a team with souped-up versions of these translation tools and the expertise and knowledge to pull it off. But small businesses that want to grow big should not write off machine translation just yet.
From Lauren Supraner It is no surprise to me that the author of “Goodbye Language Barrier?” (March) was disappointed in his translation software. Translation software will continue to have problems as long as it focuses on the language and not the message. Communication is deeply tied to cultural identity, and the two cannot be separated. What you say, the words and grammar, are what most translation software focus on. However, communication is primarily about how you say it– or pragmatics. Pragmatics includes how you influence others, or express your opinions, desires or emotions, depending on context. Translation is about the best way to get your message across, which cannot be accomplished outside of a cultural framework. For now, a human translator, who understands the nuances of getting your message across to the right audience, remains your best bet.
From Jiri Stejskal, Elkins Park, PA I read with interest Mr. Leighton’s article titled “Goodbye, Language Barrier?” in the March 2008 issue of the Fortune Small Business magazine. I heartily agree with his conclusion that translation software applications are not suitable for business use and that human translators are needed to facilitate cross-border business. I would like to share the observations below with you.
In an effort to cut costs, some businesses turn to machine translation (MT). Machine translation has been around for longer than most people think. The first proposal for using MT was submitted by a researcher at the Rockefeller Foundation in 1949, and in 1954 the first public demonstration of an MT system was held in New York at the head office of IBM. Much hope had been put into MT during the Cold War years, with disappointing results. Today it is clear that while MT is here to stay, it has limited use and will not replace human translation anytime soon. Let us look at “Free Online Language Translator” by WorldLingo to demonstrate the limitations of MT. Consider a fairly simple question that could be used in a questionnaire:
Thinking about this feature, please indicate your level of agreement with the following statements. Use a 5-point scale where a “1” represents “Strongly Disagree” and a “5” represents “Strongly Agree.”
After we paste the text into the Free Online Language Translator (http://worldlingo.com/en/products_services/worldlingo_translator.html) and select Spanish as the target text, we receive the following result:
Pensando de esta característica, indique por favor su nivel del acuerdo con las declaraciones siguientes. Utilice una escala de punto 5 donde discrepa un “1” representa “fuertemente” y “5” representa “convienen fuertemente.”
Using the same tool, but now translating the Spanish rendition back to English, we get this remarkable statement:
Thinking of this characteristic, it please indicates its level in the agreement with the following declarations. Use a scale of point 5 where a “1” differs represents “strongly” and “5” represent “agree strongly.”
Besides the obvious grammatical and stylistic errors both in the Spanish and the back-translated versions, we see that the meaning of the original statement has changed significantly. “Strongly Disagree” became just “strongly,” thus rendering the scale meaningless. The request (“please indicate…”) became a statement (“it indicates…”) with a nonsensical “please” in the middle of it. In addition, during the translation process, the user had no way of indicating who the target audience would be. In the case of Spanish, it makes a big difference whether the translation is geared toward a panel in the US, Mexico, South America, or Spain.
So is MT a waste of money? Not necessarily. There are situations for which MT is appropriate and can bring positive results. Also, there are MT tools available on the market that are far more sophisticated than the WorldLingo application we used in the example above and many other “free” programs such as Google Language Tools and AltaVista Babel Fish. SYSTRAN provides off-the shelf products, and companies such as LanguageWeaver are offering licenses for the MT engines they develop for a substantial fee. MT has been used successfully in a controlled language environment by Caterpillar and other companies. With vocabulary restricted to some 800 words and rigid syntax rules, Caterpillar’s technical writers were able to produce English documents suitable for MT, yielding remarkable accuracy.
MT is also used by the US government to scan large amounts of data in order to identify certain topics or keywords; such documents are then typically submitted for human translation. The European Commission uses its Machine Translation Service for a similar purpose, and informs the users that “[MT] cannot compete with the accuracy or quality offered by professional translators. However, it can still be helpful, particularly when time is at a premium. Machine output may be adequate, for example, when you need a quick overview of a document written in a language you do not understand.” (Source: https://webgate.cec.eu.int/mt/ecmt/html/help_en.html) It is not wise, though, to try to turn this idea around to give a ‘quick overview’ of your text to the world at large. Take for example the official site of the Philadelphia city government, which uses machine translation to translate its web pages into multiple languages. The city’s mayor, Mr. Street, becomes Herr Straße, Monsieur Rue, and Señor Calle, among other things, and the “Lead Story” becomes a “Story of Lead” (as in “lead paint”).
Evidently, machine translation is both a friend and a foe. Just like any complex technology solution, it can enhance productivity and quality, but can also do the exact opposite if not used properly. Much research and development is needed to design a good MT program capable of recognizing context and following syntactical rules, and therefore it comes with a price tag. “Free” MT programs which operate on a word level disregarding the context are a good source of office ridicule, but certainly not a tool for a business professional.
From Ken Zwerdling, CEO, Foreign Translations, Inc. I greatly enjoyed "Goodbye, Language Barrier" (March). Successful international expansion demands using professional translators, not machine translation. There is absolutely no substitute for human translators educated and experienced in fields of engineering, law, finance and medicine. Trying to cut costs on translation risks damaging your company's global reputation. These tools may increase in accuracy over time, but not for many years in the future.
From Alex, Jerusalem, Israel I was part of a development team for translation software for 4 years back in the 80s, and it eventually became pretty obvious to almost everyone involved that machine translation (as it was called then) was a no-go. It wasn't a problem of slow computers or anything like that. It was just that human language is so complex that extracting meaning from a sentence (one skill) and producing an equivalent sentence in another language (another skill entirely) is well nigh impossible using conventional algorithmic methods. As long as massively parallel (and I mean massive) neural network computing is still out there in the future, computer translation of human languages is going to sound like it was produced by a dumb machine.
From KK, New York City Spend the money on a foreign language class instead of the software. You get more out of the class than you can from the software and less headaches along the way.
From Tom, Milwaukee, WI The best results come when you use a higly usable, secure, and robust workflow within a content management system. Human translation services in the workflow are the best to hit cultural nuances. Titan CMS from Northwoods Software in Brown Deer WI has inluenced this market quite nicely.
From Val, Amsterdam Well, natural language processing is indeed a very difficult problem. From the packages I've seen, you get a very fast translation of the text, but which you still have to process yourself to make any sense of it. For an unknown language you can get pretty much a rough idea of what a text is about and the main ideas in the sentences. However, it does not replace a human translator in any way: you really need to hire someone who is an expert in the destination language to think and brush it up considerably (and even then, it's not the same performance is not the same hiring an expert translator). So bottom line: if you want to know fast what a completely foreign piece of text says (e.g. like when surfing the web), they are pretty useful. If you need to use them in business/official documents/press releases, you definitely still need to hire a human expert.
From George Smayling, Houston, TX I have used Systran. It has been a few years. Out of the box performance was okay but not at the level of a skilled human translator. The custom dictionary feature yielded better results, particularly when dealing with industry specific jargon. The client I was helping located a bilingual technical dictionary for his field and added many entries to his custom dictionary. One particularly useful feature is storage of phrases in the custom dictionary. As you may have noticed, the meaning of a particular phrase may be lost entirely when the translators act on each word separately. While using custom dictionaries did not reach the level of a human translator, Systran saved time overall. Integration of Systran with MS Office programs enabled users to quickly generate reasonably accurate translations which a human could clean up with minor editing to improve readability. Good luck with your translation projects!
From Yeah this works We like to use the software of translation of language. The results are very funny. We always are ready for a voucher to laugh, it breaks the day. Do not trust your matters to him!
From Fred Deutsch, NY, NYC I'm wondering if you've ever heard of technology by Voxonic which lets you present a speaker in any voice? The translation is done by hand because as you know translation software doesn't work but once you have the translated text you can present in the voice of the company spokesperson in every language.
I have actually used a free application called google translate, it was developed by google and I have had no problems when talking to people in france.