Did you miss a session from GamesBeat Summit Next 2022? All sessions are now available for viewing in our on-demand library. Click here to start watching.
Machine translation has come on strong in recent years, with myriad online tools and apps serving as decent alternatives to the arduous process of, well, learning a new language yourself.
A few months back, Microsoft claimed a “historic milestone” when it said it was able to leverage artificial intelligence (AI) to match human performance levels in translating news from Chinese to English. And Google brought its neural machine translation-based (NMT) translation smarts offline — Google Translate users on Android and iOS can now access high-quality translations in 59 languages sans internet.
But as good as machine translation tools are getting, many factors can collude to make relying on machine translation alone a bad move. Overly technical subject matter, unusual language pairings, and awkward source material formatting are among the elements that mean businesses — or anyone with mission-critical translation needs — are often better served by a computer-assisted translation (CAT) approach.
Humans may have to assist with many stages of the translation process, including managing documents, preparing text into a format that’s ready for translation, and — of course — translating and proofreading material. Companies will often have large segments of text, whether individual words, phrases, or paragraphs, that have previously been translated — so rather than doing all that work again, a translation memory database can ensure the translator only needs to work on new text. Naturally, this database improves and expands over time, so the more you use a particular system, the more automated future translations will become.
Intelligent Security Summit
Learn the critical role of AI & ML in cybersecurity and industry specific case studies on December 8. Register for your free pass today.
Alternatively, a company may be perfectly happy with an entirely machine-translated document, or they may be willing to pay a human to proofread and correct the machine’s errors.
Most translation agencies now follow such a setup, pooling machines with humans to optimize processes, quality, and prices depending on what the client wants. But when you throw all those elements — humans, machines, processes — into a giant pot, what you have is a recipe for confusion, which is why many translation companies use a centralized platform that connects everyone involved in the translation process.
One of those is Smartcat, a software-as-a-service (SaaS) company that automates translation workflows. Smartcat was actually developed by Abbyy, the Russian company that develops a bunch of popular document management and text-recognition tools, and was spun out as an independent entity in 2016.
This week, Smartcat announced it has raised $7 million in a series A round of funding led by Matrix Partners, a San Francisco venture capital (VC) firm with a number of notable exits to its name, including Hubspot (IPO), Oculus (Facebook), Zendesk (IPO), and Flatiron School (WeWork).
Smartcat’s platform integrates with more than 10 machine translation engines, including Google Translate (neural and statistical), Microsoft Translator, Amazon Translate, Baidu, and Yandex.Translate. The reason there are so many is that each engine comes with its own strengths, such as proficiency in a particular language combination or subject area. The automation is actually free for companies to use, but Smartcat leverages this to sell human services — including professional translators, proofreaders, and translation agencies. The platform also automates project management, linguist appointments, and vendor payouts while serving as a central conduit that facilitates collaboration between translators and proofreaders.
In many ways, language is the final frontier of a global planet, which is why so much investment is being made into helping people communicate across countries, continents, and cultures.
Microsoft already enables real-time spoken word translation in Skype across numerous languages, including Japanese, Spanish, Mandarin, Arabic, and Russian, doing so through AI techniques such as deep learning and artificial neural networks. LinkedIn recently added a new feature that automatically translates posts in your feed, similar to what Twitter and Facebook have offered for years. A few months back, news emerged that Facebook is now using unsupervised machine learning — where it doesn’t have many existing example translations to work from — to translate content on its platform.
Elsewhere, Unbabel, a machine translation platform that uses humans to verify translated content, raised $23 million from some big names in the technology realm, including the VC arms of Microsoft, Salesforce, and Samsung.
The global translation and interpreting services market is a $45 billion industry, while the machine translation industry specifically is estimated to be around $300 million, according to Research and Markets, and is expected to hit almost $1 billion by 2023.
And this is what investors are looking to tap into.
“Our first impression was that they [Smartcat] have a very unique product — allowing companies to automate the translation process and scale content into any language on-demand in just a few clicks,” said Matrix Partners general partner Hardi Meybaum. “We also liked its network component, which makes collaboration and payments between all market participants really easy and efficient.”
Smartcat claims it garnered a 520 percent increase in sales last year, supported by a network of 200,000 linguists and 5,000 registered translation companies. With another $7 million in the bank, it plans to move its headquarters to Boston from San Francisco and focus on growing its sales and marketing efforts.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.