There is much sensational information available for people out in the ether of the Internet that over explains and complicates machine translation. Additionally, the process or the concept gets a bad reputation from translators and from agencies alike for one major reason– it is just not understood.
The client who needs translations thinks “This is amazing! I will get it faster and cheaper and it will be good quality!” However, the translator thinks “This is a machine, it can’t understand the nuances of language.” Then there comes the professional fear that tells each party that he or she is going to lose money which makes the demands of each party unreasonable and causes a breakdown in the collaboration and the conversations between the parties.
Experience has shown that all MT is not created equal. Each different company or different creator of the algorithm used to fit a non-mathematical concept like language, which is constantly changing, into a mathematical concept to be utilized in a more efficient way. It is also important to note that not all MT is created equal. The technology is still new and there is no gold standard for it yet, and the gold standard may never actually come to fruition for the simple reason that language changes so rapidly.
On the one hand, MT like Google Translate is based on a good algorithm and it has great potential because of the ability translators have to fix and choose the best possible translation. Additionally, with translation programs like memoQ the option to connect directly to MT databases like Google Translate is available. This connection has the potential to make translation much faster and more efficient, but many companies do not allow for this as a result of non-disclosure agreements and privacy concerns. However, even with Google Translate, not all languages and language pairs are created equal within the MT system.
An example to illustrate this would be the great amount of content that would be available for English to Spanish for the medical device industry. Economics play an important role in this also; most medical devices are created in countries where English is the native language, or they are created in an environment where English is the lingua franca. Thus, this language pair for this industry would have more content than say, French to Dutch for the same industry.
MT also builds and creates automated translation based on, not only the algorithm, but also the submissions of human translators. This introduces the potential for human error introduction. The general thing to remember is that humans create errors and machines create errors. Neither man nor machine is perfect in performance regardless of engineering or education.
What this all really comes down to is the truth that human translators cannot be cut out of translation completely. Human impact and human editing will always be needed and the potential to make translation more consistent and more efficient is very present, but the buy-in from both client and translation provider is required. Additionally, if there are not people who speak more than one language there cannot be machines that make translations simpler and faster. There would be no reason at all then for MT to even exist for it cannot exist in a vacuum nor can it exists without human experts to develop it.
MT really is a call to the translators to develop professionally and to embrace technology and be a part of the new technology understanding that it adds to expertise. This is especially true as new economies are becoming larger and more powerful (Brazil, Russian, India, China, etc.) and the need for translation is growing. It is growing so fast that MT has had to be developed to keep up with the demand and to meet the global demand for translation.