Does Google Translate Really Work? [INFOGRAPHIC]

Apr 6, 2015
9:55 AM

Editor’s Note: Last week, Ashley Harris sent us an interesting infographic, so we asked her to share more context about it. This is the post she sent to us over the weekend. This is not a sponsored post, nor are we endorsing the company Ashley works for. We are just geeky bilingual editors who love having conversations about language and the Internet, and this topic caught out attention.

The Internet is truly a global network, bringing us data in most human languages. Usually, we browse websites in our own language. Sometimes the information we are looking for is only available in a foreign language. In these cases, machine translation tools can be a useful asset.

Reliable machine translation is a goal that computer scientists have struggled with since the 70s. It’s a deceptively simple problem—to translate words from one language to another. However, the deeper researchers have looked at the problem, the further a solution seemed, until relatively recently.

One of the most outstanding examples of machine translation is Google Translate, which has the added advantage of being free to use. The software at the heart of the service is based on the work of Franz Josef Och. As with most modern computer scientists, he favors a statistical machine learning approach to natural language processing.

The huge databases of Google, containing massive volumes of text in virtually every language on Earth, make this type of a statistical system viable. The result is a system that surpasses previous attempts, such as AltaVista’s Babelfish. But the test of any solution lies in its applicability. Is Google Translate actually useful for web users?

Google Translate is available on a broad range of platforms, including Android devices. Most users have tried it at one point or another. While the technology is undeniably impressive, it is not without its limitations. Nevertheless, Google Translate is useful for capturing the general gist of a text.

Of course, there will be times when the general gist is not sufficient. Furthermore, the quality of the translated passages can be inconsistent. Sometimes a text is translated almost perfectly, and at other times there can be flagrant errors. Incorrect translations and phrases that remain unchanged from the original language are jarring to the reader and can often obscure the true meaning of a text.

Obviously this means that the service is not appropriate for certain uses. But for other situations, it can be extremely useful due to its speed and the fact that it is free.

The decision on whether to use the service or not depends on the context—there are times when a general sense of the message is all that is required. Conversely, there are times when a perfect duplication of the original meaning is essential.

The alternative is to use the services of a bilingual human to translate the text from the original language into the target language. However, this is not always viable. For instance, a casual web user may be checking online reviews in a foreign language to see if a product would make a good purchase. This user needs immediate feedback, whereas hiring a professional translator would require time.

Understanding the strengths and weaknesses of Google Translate can help to decide when to use it and when to opt for the human touch.

In order to simplify the decision process, translation and transcription service firm Verbal Ink, conducted a series of task specific trials. The findings are presented in the infographic below.


verbal-ink