At Google I/O 2022, held on 11 and 12 May, the firm revealed that its popular Google Translate software now encompasses 24 additional languages, taking the total to 133. The most recently added languages include tongues from Africa, Asia and the Americas.
The news of the additional languages, which I’ll cover in more detail below, means that businesses using machine translation software now have more indigenous and rare languages at their fingertips than ever before. With that in mind, let’s fire up the whole human translation vs machine translation debate once more.
Can a human translate better than a machine? This is the question that many businesses ask when they’re weighing up the human versus machine translation decision. But the answer depends entirely on what we mean by ‘better’.
A human translator can deliver greater nuance and refinement, certainly. This is central to the case for human translation. Machine translation software doesn’t yet provide all those little tweaks and adjustments to the text that make human translators’ work such a skilled undertaking. Ambiguous expressions, idioms, gendered language… all of these benefit from a watchful human eye (you can read more on how to translate gender by clicking the link below).
However, the machine translation vs human translation debate doesn’t end there. Because machines can translate much faster than humans. And what cost-conscious business can afford to ignore that fact?
Read more: How to Translate Gendered Language
The potential cost savings are why there has been such a sharp increase in demand for machine translation post-editing (MTPE) services in recent years. Businesses are using machines to translate fast, then humans to refine the translations. It’s no longer a question of simply human translation vs machine translation, but rather of taking the best that each can provide and blending them. The speed of machine translation is complemented by the unique touch that human translators bring to the table.
If you’re after an in-depth exploration of machine translation post-editing, the link below should help.
The latest additions to Google Translate hail from Latin America, Asia and Africa. Collectively, they have around 300 million speakers. Businesses looking to connect rapidly and easily with speakers of these languages now no longer have to rely on human translation alone.
The full list of 24 addition languages includes:
· Quechua (around 10 million speakers)
· Guarani (approx. 7 million speakers)
· Aymara (approx. 2 million speakers)
· Bhojpuri (approximately 50 million speakers)
· Maithili (approx. 34 million speakers)
· Assamese (approx. 25 million speakers)
· Ilocano (approx. 10 million speakers)
· Kurdish (Sorani – approx. 8 million speakers)
· Dogri (approx. 3 million speakers)
· Meiteilon (Manipuri – approx. 2 million speakers)
· Konkani (approx. 2 million speakers)
· Mizo (approx. 830,000 speakers)
· Dhivehi (approx. 300,000 speakers)
· Sanskrit (approx. 20,000 speakers)
· Lingala (approx. 45 million speakers)
· Oromo (approx. 37 million speakers)
· Luganda (approx. 20 million speakers)
· Sepedi (approx. 14 million speakers)
· Bambara (approx. 14 million speakers)
· Twi (approx. 11 million speakers)
· Tigrinya (approx. 8 million speakers)
· Tsonga (approx. 7 million speakers)
· Ewe (approx. 7 million speakers)
· Krio (an English dialect – approx. 4 million speakers)
Humans who translate these languages aren’t going to be put out of business overnight by Google’s machine translation software advancements. However, it would be wise for them not to ignore machine translation either. Those who can rapidly update their business model to offer MTPE services for these languages will be well positioned to take advantage of a new wave of demand.
Businesses will have some decisions to make as well. A fundamental question of decisions around machine translation vs human translation is, ‘How good is good enough?’. For some businesses, only the highest quality of translation will suffice, in which case the finesse of a human translator will be essential. In other cases, something that is translated well enough to convey the intent of the original, without needing to be word-perfect, may be good enough.
One of the most exciting elements of Google’s reveal is the tech that lies behind the addition of the 24 languages. The firm has used ‘Zero-Shot Machine Translation’ technology, meaning that it only needed to feed monolingual text into its machine learning model. Other models need parallel (translated) text from which to learn, but the Zero-Shot MT model can deliver translations “without such data,” according to Google.
While the technology isn’t flawless yet, it carries major implications for adding new languages to Google Translate over the coming months and years. And this has major implications for companies looking to overcome language barriers in order to maximise their potential (for more on that, click the link below).
Ultimately, businesses that can save time and money without seeing standards slip have the potential to outperform their competitors. Advances in machine translation are supporting businesses to do this, particularly when human translation expertise is added into the mix.
All of this points very clearly to machine translation, and machine translation post-editing, being the way forward for the translation sector. The outdated human vs machine translation debate has been replaced by an ever-closer mixing of the two approaches, with human translators’ skills complementing machine translation software’s speed.
This new paradigm is enabling businesses to reduce the time their translations take, as well as the cost of producing them. It is also supporting them to reach out and connect with new audiences, thus opening up new revenue stream potential and new possibilities on a global scale.