Analyzing Speech Development
September 02nd, 2015
A child learning to speak the English language must master 44 different speech sounds. Some of these sounds take years and years to develop and it can be incredibly complex to determine if language is developing normally over time.
“A child or a second-language learner of any language tries to reproduce words of that language and the sounds they contain,” said Yvan Rose, an associate professor of linguistics at Memorial University.
“How do we first compare their sounds to what they should be saying, and how do we capture their evolution over time?”
A lack of accessible and affordable tools to assess child language development presented a challenge to his research, but also a unique opportunity. Software and hardware solutions at the time Rose was trying to answer those questions offered only limited functions, making side-by-side sound comparisons difficult to assess, and they also came with a hefty price tag — in the tens of thousands of dollars for some systems.
In 2004, this problem led to a novel idea that has had significant repercussions for our understanding of language development in children and new language learners, even today.
“We wanted to have an understanding of how the learning works so we have a sense of when it’s happening properly as opposed to when there are challenges. That’s why we created Phon,” said Rose.
Rose teamed up with a psychology professor from Carnegie Mellon University, Brian MacWhinney, and MUN software programmer Gregory Hedlund, to develop Phon.
Phon is a linguistic analysis software tool that analyzes speech sounds and their combinations within syllables and words. At the time of its inception, Hedlund was a computer science student. He has been working full-time on Phon ever since.
The project’s potential was realized early on. In addition to startup funds from MacWhinney, Rose received the Petro Canada Young Innovator Award (now called the Terra Nova Young Innovator Award). The award supports outstanding faculty members whose research idea is particularly novel and innovative, and has a real potential to make a significant impact on society. The award, cost-shared between MUN and Suncor Energy Inc., is a stepping stone for emerging scholars such as Rose.
Almost from its inception, Phon has played a key part in a global research consortium called PhonBank, an extension of CHILDES (the Child Language Data Exchange System), which continues to receive funding from the National Institutes of Health.
Today, the software is used by researchers, educators and speech language pathologists in more than 27 countries and has transformed our understanding and access to language development.
The database has expanded to include samples of dozens of other languages so researchers can better understand and compare speech sounds in different languages such as French, Dutch, Arabic, Cantonese, German and Japanese. And it continues to grow.
The software is completely “open source,” meaning researchers can avail of it anywhere in the world free of charge, with the only caveat being that they give back to the research community by sharing their data.
“Because of lack of data availability at the time, research questions that used to take up to 10 years to address can now be answered in a matter of weeks,” said Rose.
The tool can help researchers and professionals identify, diagnose and treat speech disorders in children.
“An earlier diagnosis is always better, because if you catch a speech issue early on, it’s much easier to fix,” explained Rose.
The tool has put MUN on the map for linguistics research.
“Yvan and his team are not only contributing to a better understanding of speech and sound development,” said MUN vice-president (research) Richard Marceau. “He’s also aiding thousands of researchers, students, educators and speech language pathologists around the globe. Phon is internationally recognized and embraced for its contribution to understanding speech.”