The rise in popularity of machine translation

The rise in popularity of machine translation

It’s the start of many memes, jokes, and laughs on the Internet: machine translations. Anyone who has tried to plug a sentence into the Google Translate tool knows that once in a while, a computer struggles to churn out a perfect, grammatically correct translation. While some errors are subtle, many are humorous. In the worst-case scenario, machine translations are culturally offensive.

With so much on the line, many wonder why a company would ever leave marketing materials, product information, or research in the hands of a computer. Yet, studies show that machine translations continue to increase in popularity.

The difficulty of machine translations

People have used machine translation for years. Before someone programmed computers to translate text in seconds, machines broke down source texts and reconstructed them using basic grammar rules and definitions. The process was tedious and time-consuming. Since then, machine translations have come a long way. Now, they are done based on numbers and historic behaviors.

By taking large volumes of data from human-translated text and feeding it to machines, programmers help the new models to “learn” how to translate texts more appropriately. The goal is to increase accuracy. The more human-translated texts that get fed through the computer, the higher the likelihood that future translations will come out error free.

Google Translate is one computer program with access to significant amounts of data. That’s why it continues to become more and more accurate over time.

The benefits of machine translation to businesses

Although relying on a machine translation for every translation might not be ideal, there are certain times when it makes sense.

  • When you need to get the gist of a letter or document

  • When you need to summarize a legal document

  • When you need to understand something on the web

Each of these and countless other scenarios has one goal: to get a basic understanding of what the source text says. Perfect grammar is not necessary in these instances.

Where machine translation does not excel

As with any automated system, there are a few downsides to using computers to translate documents. Machine translation is ideal for many jobs, but there are some places it does not excel at. These include:

  • Idioms. Some people use machine translations like an online dictionary. While this works the majority of times, idioms and double meanings to various words are difficult for a computer to detect. Sometimes this can turn around inaccurate translations.

  • Sentence structure sometimes gets confusing. As you start to translate longer, more complex sentences, grammar sometimes gets lost. Unlike short phrases, machines rarely see the same sentence stream through repeatedly. This makes it difficult for the model to learn what is correct. When it comes to sentences on warning labels and signage, it is important that a human verifies the text is accurately translated before making it public or putting it into use.

  • Paragraphs and longer text translations give the gist of the document. The more words, sentences, and structures you feed into a machine translation tool, the more disjointed the outcome gets. Grammatical errors multiply. For technical legal documents, or product descriptions where accuracy is crucial, machine translations are not ideal.

While there is some level of risk involved in machine translations, they continue to increase in their usefulness. The reason is simple. In spite of the errors, machines work fast and continue to develop so that they are more accurate than ever. They also provide a basic outline to human translators to make it faster and easier to edit and perfect the text.

For those instances where accuracy is important, machine translations are fast and easy to edit. With a post-edit of a machine translation, companies still save significant amounts of time (and money) while getting a perfect, grammatically correct translation. The machine translations are easy to read and understand, even for a native speaker, making them fast to edit.

Although machine translations might never replace human work, they continue to serve as a strong aid in the process. With the amount of time, money, and energy computer programs save translators today, it makes sense that this method continues to grow in popularity.

Image: PhotoSpin