Machine Translation: Why You Should Not Depend On It
Updated: Mar 10
Machine translation, abbreviated MT, is a software tool that can automatically translate content from different sources into the languages specified. MT software may be used by humans to help them convert speech and text from one language to another.
MT software can translate massive amounts of information, however, the quality of MT software translation varies greatly. Usually, companies offering translation services will use MT to reduce cost the cost of human labor, boost the productivity of the translators they employ, and provide editing services to their customers. However, it is still missing the human component of figurative understanding."
At Liberty Language Services, we know the value of empathy when it comes to translation and interpretation. Here’s our guide to machine translation and why it’s a tool upon which you shouldn’t always rely.
Speed vs. Accuracy
Companies and medical facilities offering language translation and related services have significantly increased the use of MT software. SDL (Simple DirectMedia Layer) is one of the largest language software companies worldwide.
In 2016, SDL announced that the MT software could translate twenty times more information when compared with the output of human teams. However, this data only accounts for quantity, not quality.
There are still a lot of factors that the machine translation AI (artificial intelligence) needs to work on to be remotely close to the accuracy of a human translator. Despite the speed of MT, we still need a human translator to edit and proofread. This means that using MT may actually take more time, not to mention waste more resources.
Are you in need of translation services? Liberty Language Services offers translation for industries like healthcare, law, government, and education.
MT or AI can’t always understand culture or slang. English is a language that has 160 distinct dialects around the world, and some words take on different meanings in each of these. Only a human translator, who has adaptive linguistic capabilities, will understand regional uses of words and the ever-growing lexicon of modern slang.
There are always new words in languages and it is easier for human translators to pick up on new languages or words. However, it's difficult for machines. They will require new programs, software, and algorithms to help them. That means the machine will have to be kept up-to-date regularly.
Words in Context
Some words have different meanings and parts of speech. A machine translation might understand a verb as a noun that means something completely different.
Style and Tone
Words can signify an alternate meaning based on the style or tone of a conversation. For example, if someone says "that sounds great" in a sarcastic tone, the machine won’t be able to convey the double meaning. It could be translated positively when, in fact, the person is angry or frustrated.
Related: Liberty Language Services
Most translation companies evaluate the quality of the MT using the additional effort required by a human to edit the output. The number of keystrokes or pages completed per hour indicates the effort required. Automated tests and metrics are used by specialists who are trained in using the software since even the smallest changes are accurately recorded.
Bilingual evaluation understudy, abbreviated BLEU, is the primary metric used in auto testing. It compares the output for human and machine translation, with one corresponding to high quality and zero corresponding to low quality.