Machine Translation: A valid alternative?
When translating their websites, many organizations choose a path that, at first glance, seems to be simple and affordable: machine translation. However, as tempting as this alternative may seem, using it without the proper investment of time and money can lead to a result that is —at best— uncertain.
How efficient is machine translation for website content?
What is machine translation?
It is software that has been around in the linguistic services sector for some years, and has improved steadily as technology moves forward. The first versions had a dictionary and applied the grammatical rules of the target language. In time, new functions were developed and 3 types of machine translation were developed:
- Based on linguistic information.
- Based on statistics (Statistical Machine Translation).
- Based on neural networks (Neural Machine Translation).
Beyond the specific technical features of each and the huge investment big companies (such as Google) make in R&D, this tool is yet far from being widespread.
If we consider health care scientific texts, only 5 to 43% of papers translated with machine translation were considered accurate.
Source: National Center for Biotechnology Information
Machine Translation: Speed vs. Accuracy
Many free solutions promise “accuracy” in their translations, but without the support of a professional linguist to assure the quality of the translation, we would be sharing content of untested quality. These are some of the limitations of machine translation:
● Accuracy:
Translating is not simply replacing a word with another: it requires a complex process to understand, analyze, provide context for, interpret and reformulate ideas. Even with state-of-the-art technology, no software currently available can understand context when dealing with certain phrases and words.
● Ambiguity:
These tools are sometimes unable to retrieve the precise meaning, particularly with ambiguous terms that may hinder reading. Moreover, since they not always have enough information, they may use alternatives and synonyms that are inadequate.
● Use Limitations:
Since machine translation follows systemic structures, content customization is not possible. Another limitation is that they do not always support all file formats, such as PDF, TXT or JPG.
● Adaptation:
Some texts require adjustments for the target culture or audience; each region and market have specific characteristics, such as dialects, expressions, jargon and variants. If this aspect is not considered, the translation may even use words that could be offensive for a certain audience.
● Post-production:
Once the translation is finished, it needs to be reviewed and corrected: the more complex the material —like scientific and technical documents—, the bigger the likelihood of errors with the tools currently available.
Although both machine translation and digital transformation are here to stay, we need to be aware of their limitations. One example of this was seen at the Boao Forum in April 2018: Tencent –a multinational technology company– introduced its machine translation tool based on neural networks and, after a demonstration showing its functions, the tool made several mistakes, like repeating words and misusing phrases.
The accuracy of English to Arabic translation in Google Translate is only 57.15%.
Source: Scientific Research Publishing
We cannot deny that this technology is very practical to translate simple texts for immediate use or to understand what a text is about. But when we are trying to expand our business and reach new markets, this poses a risk to our corporate image, because a poorly translated text makes the audience question how professional our company is.
It is advisable to invest in the quality of content that has a strong bearing on our brand, with providers offering professional services that are customized to each region, field, culture and sector. How can we find such a provider? Quality is the bridge that will allow us to reach our audience and develop long-lasting relations.