The world is getting smaller, businesses are getting bigger, and individuals are recognizing the value of looking abroad for talent, collaboration, and success.
The past showed businesses dozens of reasons to stay domestic. Most of these reasons have been made irrelevant by ever-improving technology. Heavy tariffs lose relevancy in a digital world. High cost of travel doesn’t exist with cheap flights and video conferencing and communication tools like Skype, Slack, and Workplace by Facebook.
But the most important barrier, language, has remained a decisive limiter to the true potential of global expansion. How can a US based company do business in Vietnam without speaking Vietnamese? How does an Estonian entrepreneur raise money from Silicon Valley investors without communicating her vision in English? And how can a small German business hire developers from Colombia without speaking Spanish?
Until recently, this sort of ambitious international expansion required expensive multi-lingual human talent (a niche owned by Benny the Polyglot, who speaks 7 languages), translators, and a whole lot of trust.
But due to recent developments in Natural Language Processing (NLP), this final barrier is about to come down. This will create enormous opportunity for global business expansion and easier business travel.
What is Natural Language Processing?
According to the definition given by Search Business Analytics, NLP is a computer’s ability to understand language as it is spoken. That essential asterisk – as it is spoken – is what makes development in NLP so exciting.
NLP is an essential instrument in enabling features like voice-to-text dictation, as highlighted by Microsoft Garage’s recently launch and promotion of a dictation app. It’s used in any voice AI, like communication with Siri, Alexa, or your Samsung-deployed talking refrigerator. In fact, this ‘search’ feature (a human asking a computer for information) is where most of NLP research is focused.
From the Sabre Emerging Technology in Travel Report:
“The destiny of search is to become the Star Trek computer,” says Amit Singhal, SVP of engineering for Google Search. “You could ask it a question and it would tell you exactly the right answer, one right answer—and sometimes it would tell you things you needed to know in advance, before you could ask it.”
(If you want to learn more, we covered the power of Voice AI in a previous post.)
And when performing at it’s best, NLP can be used for on the go translation. But before we get to that exciting future, it is important to recognize NLP’s current limits.
The limitations of Natural Language Processing
In early human-computer spoken interaction, humans were required to speak in basic, structured sentences. The computer would interpret the inputted spoken phrase and match it up with a phrase in its database. Then, if all went well, the computer generates the corresponding output, as determined by its database.
The limitations of this system are obvious. Humans don’t speak in structured sentences. At best, we babel in generalities and vague meanings. So NLP’s practicality in any natural setting- like translation while traveling – was severely limited.
Additionally, there is a lot more to human vocal communication than just the words. So much of language is subtlety. Hearing your friend say the words “I’m happy we broke up,” while watching him cry creates a contradictory situation in which language, social cues, and empathy need to work together to dictate our response.
Social context is essential with understanding language. When translating from one language to another, even more complicated factors come into play.
As Ed Bussey writes in a post for Venture Beat, “I’d question how soon AI can be taught to understand political, historical, and cultural contexts in a way that would safeguard against awkward errors.”
Even if a computer were to get the words right and understand winding sentences, there is an entire sect of language without direct translation. How would a computer be able to translate, on the fly, the Czech word prozvonit, which means “to call a phone and let it ring once, to get the recipient to call you back.” Or the Danish word ‘hygge’, which translates to “a quality of cosiness and comfortable conviviality that engenders a feeling of contentment or well-being.”
Language is hard. Getting a computer to understand, translate, and respond in natural human language is a big research task.
Current Research in Natural Language Processing
While there is a lot of deep learning and development research happening in the world of NLP, much of it happens in a few key areas. David Orr, the project manager for the Google Research NLP groups, describe the segments in a Quora post.
Syntax: Speech tagging and parsing in 60+ languages. We have multiple taggers and parsers, some of which are application-specific, and which make different speed/quality tradeoffs. (ie. Defining grammatical rules and analyzing speech trends like slang, shorthand, and formal speech.)
Semantics: Recognizing entities in text, matching those entities against our knowledge graph where it’s possible, labeling the entities in a variety of ways, analyze coreference to figure out what words or phrases refer to the same thing, and so on. (ie. Solving the problem of context in speech, and being able to understand the subtleties and cultural uniquities, and adding emotional empathy and cognition.)
Knowledge extraction: Learn relations between entities, recognize events, match entities between queries and documents. (ie. Congregating information from many sources on similar topics to develop better understanding of an issue, and providing better information to humans.)
Summarization: Figure out the topics of a page, and generate summaries of the page. Sentiment analysis, clustering on a variety of metrics, etc. (ie. Understanding large chunks of text and speech well enough to generate summaries, saving humans time.)
Question answering: When is a query really looking for a piece of specific information and how do we find and surface that piece of information?
All of these research areas come together to develop Natural Language Processing’s capabilities.
Natural Language Processing for Travel
All the limitations aside, researchers are clearly making an enormous amount of progress in the space. Led by companies like Microsoft, Baidu, Google, and Apple, the NLP market is set to grow from $7.63 billion USD in 2016 to $16.07 billion USD by 2021, according to Markets and Markets.
This means that lots of progress is being made. The earliest modern developments in translation came in the form of Google Translate. Since launch in 2006, it has grown to support over 100 languages and over 200 million uses a day. This huge amount of data, combined with recent improvements in ‘neural machine translation’, make Google Translate one of the most important tools for rapid translation.
In a previous 30STF post, we described language learning and translation tools to use when preparing for a business trip. These include services like Open Culture and Memrise for language learning, and Photo Translator and the many business phrasebooks available for quick translation on the go.
The next level of that is an advanced NLP Artificial intelligence able to translate language as fluidly and fluently as possible, all in real time. (Early iterations exist in text, as discussed in our article about chatbots.) Some solutions already exist.
Google’s newest product, the Google Pixel Buds, were announced in early October 2017. They incorporate NLP in a pretty neat way. Skift describes how the Buds work for translation. A listener can hear inputs in one language, have it translated into their native tongue, respond in their native language, and have the Pixel Buds translate back in the language of their choice. The drawback is that they buds require a mobile connection with your Pixel phone, which limits the usage to data or wifi enabled areas. (Click here for a video demo.)
Another tool focused on translation is Translate One2One, a $279 IBM Watson-powered earpiece capable of translating across 9 languages, including English, Arabic, and Chinese. So while the United Nations only has 6 official languages that cover ~2.8 billion people (according to the UN), Translate One2One provides software the expands that potential usage pool immensely. This AI powered earpiece is not reliant on any connection, so can outpace the Google Pixel Buds in the scope of its usability.
The benefits of tools like this are endless. As quoted in Digital Trends, Translate One2One founder Danny May said:
“As the first device on the market for language translation using AI that does not rely on connectivity to operate, it offers significant potential for its unique application across airlines, foreign government relations and even not-for-profits working in remote areas.”
So what does this mean for businesses?
So we know the challenges of NLP, and its potential. We appreciate that technology to do automatic and fluent translation is near, if not already here. This means big changes for businesses.
By being able to use tools to ease translation and virtually remove language barriers from the work place, small and medium businesses will finally be able to confidently hire international talent. This comes with lots of benefits including a larger hiring pool, new cultural insights, and a bigger market for your product.
Growing your team internationally means more international travel. While communication tools like Skype and Slack have been invaluable in facilitating internal communication for companies, in person meetings continue to be the best way to have meaningful debate, collaboration, and discussion.
Using NLP as a travel assistant can help your best sales people pitch clients in entirely new markets and grow your business. While international sales might traditionally be done over email (to allow for text translation) or by hiring local sales people with less training, NLP technology can permit important and seamless in-person sales meetings with representatives speaking multiple languages.
This is an immeasurably important benefit. Forbes writes, “You are twice as likely to convert prospects into customers with an in-person meeting. The likelihood of getting a “yes” increases, because it is so much easier to say “no” in an email or on a phone call.”
Natural Language Processing is one of the most interesting and deep research segments in modern artificial intelligence technology. While many limitations remain, its impact on business travel and transactions is undeniably imminent.