By Rachel Wheeler
For years, Google Translate has been a language translation powerhouse. Spanning 103 dialects, the system has been a boon for businesses with cross-border markets requiring marketing translation services and other interpretation needs. The applications that make it run smoothly include 150 artificial intelligence (AI) programs which translate from one language into another, with English as the ‘middleman’.
Artificial intelligence is a broad category of computer applications intended to mimic human thinking in such a way that a human would not detect it is unnatural. Examples of AI that have become accepted technologies include voice, speech and facial recognition, chatbots and driverless cars.
One approach to AI has been a form of programming called machine learning. Machine learning gathers a vast amount of data, categorizes it in a programmed manner and returns a relevant response. The process is a statistically predictive approach. The more machine learning analyzes, the more refined its response is to a question. The “learning” portion comes into play since the programs themselves learn to identify relevant information as time goes on.
As advanced a process as machine learning is, challenges still remain for language translation. Machine translation takes an objective, linear approach, matching words on a paired basis and attempting to pick the right pair.
Unfortunately, the process lacks nuance and interpretation which leads to faulty translations. A program might, for instance, not understand the subtle difference between a “racer” and a “driver,” or miss the distinction between “tap” and “touch.”
Neural networks leap beyond machine learning
Enter Google Neural Translation Machine (GNMT). GNMT is a form of next level machine learning called neural networking. Instead of a linear approach, a series of interconnected and layered nodes form a complex program.
It is built in a way that mimics the structure of the human brain and its natural neural network of neurons and synapses. Computer neural networking is more nuanced, taking in information from multiple sources to develop an understanding of language that reflects the way people use it.
As neural networks gather more information, the correct results produce stronger connections between nodes and weaker links for less likely ones. This process, known as “training,” represents a mode of learning for machines that has rapidly become more sophisticated.
GNMT achieved a stunning result in 2017 that will have profound implications for language translation in the future. To date, most translation programs use English as an intermediary step.
To translate French to German, for instance, the software translates French into English, then English into German. GNMT seems to have eliminated that middle step all on its own.
Called “zero-shot” translation, this new process is able to take two languages it has never translated between before and proceed without that middle step. This has demonstrated that GNMT has a much deeper understanding of languages than ever before.
The future for language service providers (LSPs)
As more products like GNMT come on the market, LSPs will be able to expand and focus their services. Rather than be replaced by machines, LSPs will still provide expertise in cross-border marketing translation, business regulations and local culture understanding that cannot be replaced by algorithms. Even with current advances, software cannot replace or innovate in style or approach like the way that humans can.
What neural network translations will do is provide a means for more efficient and faster responses to clients. With the help of LSPs, clients will be able to focus on business objectives and goals rather than simple translations.
Neural networks are more than language translation
For investors interested in neural-network-based translation, a key point to remember is that language translation is just one expression of the abilities of neural networks. Language recognition, including audio and textual, are key components of advancing technologies.
Microsoft (NASDAQ:MSFT), Google (NASDAQ:GOOGL) and Apple (NASDAQ:AAPL) all have research divisions dedicated to AI-driven solutions. Cortana, Google Home and Siri are well on their way to bringing together neural networking and language recognition together. Facebook and Amazon also have their hands in developing similar neural networks.
Opportunities in the neural network field are not limited to the language translation services either. Supporting technology, like new, improved computer-processing chips, also represents investment opportunities in companies like Nvidia (NASDAQ:NVDA), IBM (NYSE:IBM) and, of course, Intel (NASDAQ:INTC) and Google.
With all of the rapid advancements being made in the translation vertical, investors should be the first to embrace these technologies to keep their competitive edge in an increasingly globalized market. As AI technology becomes more widespread and accessible, companies that optimize for it early will be able to reap the rewards.
Don’t forget to follow us @INN_Technology for real-time news updates!
Securities Disclosure: I, Rachel Wheeler, hold no direct investment interest in any company mentioned in this article.
Editorial Disclosure: The Investing News Network does not guarantee the accuracy or thoroughness of the information reported in contributed article. The opinions expressed in these interviews do not reflect the opinions of the Investing News Network and do not constitute investment advice. All readers are encouraged to perform their own due diligence.