2018 SPEECH LUMINARY: MARIE METEER, PH.D.
We present the thinkers and innovators who are creating new tools and approaches for speech technology—and fostering the next generation of talent.
Meet the ‘Speech Teach’
Speech recognition, natural language processing, dialogue management, language analytics, and machine learning can be difficult classroom concepts for beginners in the field to grasp. Fortunately for many Brandeis University students, they’re under the trusted tutelage of Marie Meteer, Ph.D., associate professor of the practice of computational linguistics in the department of computer science. With her extensive repertoire, Meteer is there to steer scholars straight.
In her computational linguistics, machine learning annotation, speech recognition, and dialogue design classes, Meteer imparts more than the textbook basics. She challenges her pupils to ask deeper questions. “I find language fascinating,” she says. “For example, why do we order a ‘large Coke’ but not a ‘big Coke’? How does our knowledge and interaction with the world affect our understanding and generation of language? There’s so much to know.”
Meteer’s inquisitive nature derives from an impressive background, which includes 20 years spent at BBN Technologies, a firm credited with helping to invent the internet. There, among other duties, she conducted ample research on language modeling and led the commercial speech team in developing interactive voice response (IVR) applications and text analytics for call centers.
“We were part of several large government research programs to improve speech recognition across different projects for telephony, broadcast news, and multiple languages, continually pushing down the word error rate on increasingly challenging speech recognition tasks,” she says.
Meteer is endlessly fascinated with the complexities of linguistics. “Speech recognition is just the preprocessing. The real problem is understanding language in context,” she notes. “A relatively clear recording of speech can be identically and accurately transcribed by several different speakers of the same language. But if I asked them to write down the representation of the meaning of the spoken recording, how would each of those transcribers respond?”
Put another way: “Language cannot be reliably interpreted out of context; and even in context, the different people might not understand an utterance the same way.”
Meteer anticipates groundbreaking advances in speech technology arriving in the near future, including better modeling of accents using acoustic models with a multilingual base and continued improvement in microphone technology.
“And from the back end, I’m looking forward to better integration with models of context to guide language models and understanding,” adds Meteer. “I’m also thankful that, after a decade of stagnation, speech technology has gotten much better and we can focus on making dialogues more natural.”
But she won’t just be teaching about this progress in the hallowed halls of academia; Meteer also plans to expand her independent consulting services and advise clients about the latest speech tech innovations and options. Current projects involve consulting for McLean Hospital, where she’s using machine learning to extract info from patients’ health records to improve outcome predictions; Partner’s Healthcare, for which she’s aiming to improve the accuracy and quality of dictated medical documents; and NavCanada, the beneficiary of Meteer’s technical support on the use of speech recognition in air traffic control training and simulation software.
“I also work with AVOKE—a call center analytics platform—on multilingual modeling and topic modeling. And I’m on the board of AVIOS, the Applied Voice Input Output Society,” says Meteer, who, in her free time, talks the language of plants as a self-described “guerilla gardener” in her backyard.
At a time when the pace of change in speech technology evolution and adoption seems to be on hyperdrive, a few key trends are pointing where the industry will move in the coming months.
We present the thinkers and innovators who are creating new tools and approaches for speech technology—and fostering the next generation of talent. In this installment, we talk to Ilana Shalowitz from Wolters Kluwer Health.
We present the thinkers and innovators who are creating new tools and approaches for speech technology—and fostering the next generation of talent. In this installment, we talk to Daniel Povey from the Center for Language and Speech processing at Johns Hopkins University.
A Game Platform Helps Nonverbal Children Find Their Voice.
Both solution buyers and providers need to stay focused on changing alliances and data security.
These cutting-edge vendors are leading the way in AI, analytics, natural language, smart speakers, and more.