Found in Translation
As Pittsburgh companies gain international acclaim in language technology, they are changing how we communicate and learn.
Photography by Becky Thurner Braddock and Martha Rial
When PayPal was exploring ways to accelerate its rapidly expanding global business in 2011, the mobile commerce company turned to Pittsburgh. With 12 million customers around the world paying with 25 currencies, the e-commerce giant needed accurate and articulate technology to translate its business for many markets. And it needed it fast.
Enter Safaba Translation Solutions. The Carnegie Mellon University spinout in Squirrel Hill, founded in 2009, uses computers to translate corporate websites, sales materials, emails and social-media platforms for multilingual enterprise. Safaba’s custom translations assist Fortune 500 companies in capturing the nuances of business messaging and mission.
The volume of written content on the web is exploding, says Alon Lavie, Safaba co-founder and research professor at Carnegie Mellon’s Language Technologies Institute. Demand is rising for technologies that beat the speed of human translation and preserve each company’s branding and vocabulary, he adds.
“Most other providers are trying to build a single system that translates as generally as possible,” says Lavie. “We focus on generating specifically targeted translations for each client individually.”
Alon Lavie, Safaba co-founder and research professor at Carnegie Mellon University’s Language Technologies Institute.
Safaba’s success is but one of many gains in a flourishing local language-technologies arena that in many ways mirrors Silicon Valley’s ascent in the 1980s as the startup capital of the world. Almost overnight, Pittsburgh has become internationally recognized as an innovative leader in developing computer-generated language software and a hub of companies changing how we learn.
Nine promising language-related tech firms are based in or near the city — Duolingo, Safaba, M*Modal, Carnegie Speech, DynaVox Mayer-Johnson, Cepstral, Kextil, WeSpeke and LightSide. Three other companies born here — Carnegie Learning, Quantum and Think Through Math — tap similar technologies to offer online learning in math and science. Each fulfills different needs, but overall, they are breaking verbal and geographic barriers while building multilingual communities that convey ideas and transact business with ease.
Mobile Technologies, creators of the mobile translation app Jibbigo, also got its start in Pittsburgh. The CMU spinout was one of the first companies to build a speech-to-speech platform that instantly translates the spoken word into text and speech in several languages. Facebook liked it so much that it acquired the firm in August for an undisclosed amount and swept it off to Menlo Park, Calif. The social network most likely will keep a Mobile Technologies lab here, observers say.
“There’s a unique opportunity for Pittsburgh to establish itself as the crossroads for language and culture,” says Michael Elchik, founder of Green Tree-based WeSpeke, a teleconferencing platform that combines personality-based algorithms with language-learning tools. WeSpeke matches people in different countries who share similar interests; together, they learn each other’s vocabulary, pronunciation and culture in real time, and use each other for practice.
“You hear semiconductors, [and] you think San Francisco. Finance, you think New York and Chicago. We need to tell this story,” says Elchik.
Michael Elchik, founder of Green Tree-based WeSpeke
The strength of this growing cluster is a result of research coming out of CMU, a powerhouse in the computer-science world. The university was one of the first in the world to make computer science a stand-alone discipline in the ’60s. Since then, its work in computerized learning and translation has blown the lid off artificial intelligence, giving rise to new fields of study.
The Language Technologies Institute at CMU was established in 1996 under the direction of Jaime Carbonell, a professor whose work in the development of natural language-processing tools and technologies has made him somewhat of a guru in the field. LTI was the first institute to study the emerging fields of automated linguistics, machine translation, speech recognition, text-mining, computer-assisted language learning and intelligent language tutoring.
“It was an opportunity to go where no one else was going,” says Carbonell. “We were creating a trend. It was a matter of being early and attracting the best human talent to do it.”
The impact of this research owes much to innovation in machine learning, a term that describes the work computers do when they compute massive quantities of data in seconds — a feat that would take us hours, if not days.
The human brain works by interpreting well-defined nuggets of information, explains Carbonell. By contrast, computers exploit data all at once, detecting and correlating patterns and making statistical inferences at a rapid pace.
Pittsburgh’s companies have built on this growing field of research, attracting some of the brightest scientists in the world to LTI and the region. Among them are Alex Waibel, founder of Jibbigo; Raul Valdes-Perez, founder of Vivisimo, which IBM acquired in 2012; and Safaba’s Lavie.
Luis von Ahn of Duolingo is changing how the world communicates.
At 34, Luis von Ahn is a rising star among them. The Guatemalan entrepreneur and CMU associate professor has amassed a list of accolades that go on much like his algorithms, including a “genius grant” from the MacArthur Foundation.
He made headlines when he sold his company reCaptcha to Google in 2009. reCaptcha is hard to miss, with its ubiquitous and often-annoying box that requires individuals to decipher and precisely type several distorted characters before a transaction can be completed.
von Ahn believes that computers should harness human energy to accomplish more than one task at a time. Some call it crowd sourcing; he calls it “human computation”— a two-for-one function. In reCaptcha’s case, each time people identify distorted letters in a reCaptcha box to prove they’re human, the typed letters are reused for Google Books, a massive international project that converts out-of-print books to digital versions.
“Why master one task when several can be achieved at the same time? My goal is to give computers tools that they aren’t able to achieve on their own,” says von Ahn. “To use it for the good of humanity.”
With one success behind him, he has raised the stakes with Duolingo, a mobile app launched in 2012. Duolingo not only teaches people to speak and write a foreign language, but it’s fun and free. Having grown up a privileged child in a poor country, von Ahn says Duolingo offers learning opportunities to people in nations such as his native homeland of Guatemala; it’s a way to give back, he says.
Like reCaptcha, Duolingo is a twofer. As millions of users run through its language lessons, they simultaneously translate web content into every major language in the world. The translations, which Duolingo believes are as accurate as professional versions, are sold for a profit to companies that need the web content.
Duolingo’s user base was 9 million at press time, and it continues to grow. Mashable called it “One of Eight Startups to Watch in 2013,” and BuzzFeed listed it among the “Top 18 Apps Every College Student Should Download Right Now.” It is also ranked No. 1 in the Google Play store for free educational apps on Android; language software giant Rosetta Stone pulled in at 40.
Duolingo’s potential hasn’t been lost on those who know a good thing when they see it. The startup raised $18.3 million in 2011 from several high-profile investors, including Ashton Kutcher and Timothy Ferriss, author of The 4-Hour Workweek.
“Why would anyone want to pay $500 for Rosetta Stone when they can learn a language just as effectively for free?” asks von Ahn. “Their mistake was charging too much and not going digital early enough.”
Rosetta Stone, for its part, spent the last year catching up. With sales plummeting, the company has had several highly publicized layoffs. It also acquired Seattle-based Live Mocha, an online language network with an international reach, and education-technology company Lexia Learning Systems, which is headquartered in Boston.
Juergen Fritsch, chief scientist and co-founder of M*Modal
Down the street one block from Safaba in Squirrel Hill is M*Modal, an electronic, clinical documentation services firm that provides voice-to-text technology for medical professionals. When it merged with Tennessee-based MedQuist in 2011, it became a 12,000-employee operation.
M*Modal’s software allows doctors to speak into hand-held recorders, capturing patients’ narratives throughout the day. Technology does the rest, generating an electronic record that is then reviewed and verified by a group of medical editors.
“The government is pumping a lot of money into health-care [information technology],” says Juergen Fritsch, the company’s chief scientist and co-founder. “At the same time, doctors are being asked to move to electronic health records. They need our tools to make that transition.”
So how many startups does it take to make a thriving hub?
There can’t be too many, says von Ahn, who believes what’s good about a region for one company is good for all. “The top people in the world are coming here. If it doesn’t work out for them [at] one place, they’ll have someplace else to go.”
“It’s a small-scale version of what happened in Silicon Valley,” adds Carbonell. “We’re not there yet, but we’re heading that way.”
You Should Know
Here are some Pittsburgh companies that have helped to grow the city’s international reputation as a leader in the exploding language-technology sector:
Carnegie Learning is reinventing the way students learn math. Its education software tutors students in grades 6-12 and post-secondary schools. Founded by cognitive science researchers and computer scientists from Carnegie Mellon University with an assist from veteran math teachers, it was acquired in 2011 by Phoenix-based Apollo Group Inc. Based in the Frick Building downtown, Carnegie Learning has a staff of 98 in Pittsburgh and 141 nationwide.
Carnegie Speech pioneered technology that teaches people how to speak and understand new languages through personalized language analysis and instruction. The downtown firm caters to professionals whose jobs require a finely tuned ability to pronounce words with native fluency, including air-traffic controllers, educators, business people and government workers. Carnegie Speech employs 14 and expects to hire in the coming year.
Cepstral provides voices to the world for a variety of industries. Established in 2001, the 10-person South Side firm creates software versions of the human voice for use in everything from medical alerts and announcements on college campuses to the automated message heard in air-traffic control systems.
Duolingo takes learning a foreign language to the mobile level on iOS and Android devices, with an effective, free app that translates web content into multiple languages. Its explosive growth potential has prompted an expansion of its Shadyside digs. The company employs 27 with plans to double in size in the coming year.
DynaVox Mayer-Johnson has made a profound difference in the lives of patients suffering from severe speech, language and learning disabilities. Its Eyemax System is the only technology of its kind, enabling patients to communicate with the blink of an eye and simple head movements. Based on a CMU student project, DynaVox was started in Pittsburgh in 1983; it has 200 local employees and 350 worldwide.
Kextil’s advanced speech-recognition technology allows field technicians to collect large amounts of data and relay it to companies directly and seamlessly through voice commands. Customers include global enterprises in advanced manufacturing, semiconductor and medical equipment. The company employs five people, two of them full-time, and it expects major growth in 2014.
LightSide uses machine learning to help educators assess student and professional writing in academic and business settings. A CMU Project Olympus startup, the Shadyside firm received $25,000 from the Gates Foundation this year. A basic product will be launched in one year, but language-assessment heavyweights CollegeBoard and McGraw Hill are already knocking on the door. The company has seven employees and is hiring.
M*Modal is rapidly expanding in one of the hottest markets today, assisting doctors and hospitals to digitally chronicle patient care with the help of hand-held recorders and speech-recognition technology. Acquired by MedQuist in 2011, M*Modal employs 200 at its Squirrel Hill office and 12,000 total; many staffers work virtually as medical editors.
Mobile Technologies has generated buzz since its ground-breaking technology, Jibbigo, was unveiled in October 2009. Developed under the direction of founder Alex Waibel, a professor at CMU and Karlsruhe Institute of Technology in Germany, Jibbigo was the world’s first speech-to-speech translator to run on a mobile phone with and without an Internet connection. The company had a technical team of 16 prior to its purchase by Facebook.
Quantum in Murrysville works on digital learning and assessment solutions for the education industry in a trio of categories: accounting, mathematics and science. Its technology is adaptable and can be integrated with textbook publishers’ content needs. The company has 25 employees and consultants.
Safaba Translation Solutions provides machine-translation solutions for some of the world’s largest global enterprises. The technology accurately adapts to each company’s language and branding, automating the translation of all digital content: websites, marketing materials, software products, technical documentation and customer communications. Safaba, headquartered in Squirrel Hill, employs 12 people.
Think Through Math, formerly Apangea Learning, is a downtown-based company offering web-based math instruction programs for levels ranging from grade 3 through Algebra 1. At press time, the company has 120 staffers but is expanding.
WeSpeke is the eHarmony of language learning. Its social-media platform and learning tools match students with language learners in foreign countries, allowing them to share and explore similar interests. A CMU spinout, the Green Tree company employs six locals who are assisted by 15-20 global consultants.