Nothing generates a hot debate faster than the topic of Machine Translation (MT). People love it or hate it, welcome it or fear it, use it or avoid it like the plague. For the typical large corporation, MT definitely has a role to play and like any other tool, if used properly can provide substantial benefits. MT, used improperly, can do lasting damage.
I like to relate MT to medication and often state it should have a label, “For Internal Use Only.” What does “Internal Use” mean? “Internal Use” is when you have content that is coming from outside the company, whether in the form of business intelligence or foreign patent searches, and you are using it to support research with no intention of publishing the material. If you plan on publishing material to the general public or making it available on a public facing venue, such as a web site, that is “External Use”. For publishing documents to a larger audience where your reputation or image is at stake, you would want a human involved, preferably a professional translator or Language Service Company (LSC).
Popular “free” MT tools, such as Google Translate, were built as a means of making larger amounts of content available to the world in as many languages as possible. These “free” online tools evolve over time but were never meant to provide “Human Translation” (HT) quality. At best these free MT tools provide a close approximation to the actual translation and in the worst case, are completely wrong, even though the translation provided may read well.
So how can a company doing research or bringing information in-house rely upon a tool with a hit-or-miss record? The answer lies in the approach and application of the translated information. The allure of MT is its low cost and rapid delivery. The old saying truly applies, “You can have it fast, good or cheap. Pick two.” When you combine a subject matter expert (SME) with a translation that contains enough information to provide an accurate “structure” of the contents, the SME can fill in the gaps with their own knowledge to complete the picture. This works well in time critical situations where decisions have tight deadlines and some information is better than none. Perhaps you need a “quick and dirty” translation for tomorrow, but want a more polished version by next week.
In the case of doing Intellectual Property research, a patent search for prior art may yield tens or hundreds of patents, many from Japan, China, Germany or Russia. The abstract of the patent may be in English, but in many cases the entire patent is not, so the researcher is left with a decision, “How much of the information is relevant and what will it cost to translate it all?”
Given budget constraints, even doing a small number of patents by HT is expensive, so MT is the only logical choice to do more research within the same budget. Ideally the researcher would like to apply the precision of HT, but only to those portions that really matter, such as the patent Claims and Examples. Similarly, for technical articles published in different languages, the entire article, while interesting, may not have the value that a specific portion contains. The ability to have the entire article translated by MT, with just pages 3-5 translated by HT means you gain surgical precision in how you optimize your translation budget.
There are other options, such as providing a post-edit to the base MT, which can improve the quality while not greatly increasing the overall cost. This option takes more time, but less than an HT and results in a “Machine Aided Translation” or (MAT). Much depends on the budget, time constraint and criticality of work being done as to which tool is best for the task. Simply having a choice of different levels of quality, cost and turnaround time is a concept that applies in other industries but is not often found in the language industry.
The pace of global business is increasing all the time. Being able to make quicker decisions based upon more information from all sources is a benefit. The ability to apply a variety of tools to optimize the quality level, turnaround time and cost of that information stream is something that benefits every corporation.Share