From Tunisia to Canada: Mitacs intern works to improve voice recognition technology
Semah Aissaoui’s journey with Mitacs began in summer 2018 as an undergraduate in Tunisia when he first came to Canada as a Globalink research intern. He progressed on to receive the Mitacs Globalink Graduate Fellowship and attend graduate school at Polytechnique Montréal in Canada with the award. He now works with Fluent.ai on a Mitacs Accelerate award, researching how to remove background noise and reverberation from sound signals to enhance the accuracy of offline voice-activated devices.
As an engineering student at the National Institute of Applied Sciences and Technology in Tunisia, Semah combined his desire to learn with his passion for contributing to something that will benefit society. While researching his options, a colleague told him about Mitacs, and he was immediately impressed with the array of research projects available to international graduate students.
From Globalink to Accelerate
Aware that Montreal is an active hub in the field of artificial intelligence, Semah applied to the Mitacs Globalink Research Internship program and began his work in robotics at Polytechnique Montréal in 2018. During this internship, his interest in artificial intelligence grew. He began to seek out a supervisor for his master’s degree along with a research project that would accelerate his learning and competencies in this continually advancing field.
Through the Mitacs Globalink Graduate Fellowship program, Semah applied to work with Professor Antoine Saucier at Polytechnique Montréal. He started his master’s program in January 2019. Mitacs Globalink Graduate Fellowship funded the first eight months of his program.
Professor Saucier was already working on a project with industrial partner Fluent.ai funded by the Mitacs Accelerate program. He proposed to Semah to join this opportunity.
Semah began his Accelerate internship in September 2019 and is grateful to have the opportunity to work with both Professor Saucier and Fluent.ai — a company well known in the field of speech recognition.
Voice recognition research applied to industry challenge
Fluent.ai develops speech understanding solutions for offline devices. Unlike devices such as Siri or Alexa that interpret vocal commands in the cloud, Fluent.ai’s solution can perform speech recognition directly on edge devices. In addition to providing high accuracy speech recognition, Fluent.ai is committed to developing systems that are very low-footprint and can operate on-device and entirely offline, which addresses many privacy and security concerns.
In this phase of the project Semah uses his computer engineering background and his command of programming languages. He is now coding with Python and is learning to use new machine learning libraries. Semah’s research focuses on cleaning and eliminating background noises and reverberation from sound signals that microphones collect on voice-activated devices.
This is crucial for the accuracy of speech recognition particularly when the user might be at a distance from the microphone or in a noisy environment. Cleaning the signals will greatly enhance the functional performance of voice-activated solutions in a wide variety of settings.
Work-integrated learning experience
Semah is thrilled to work alongside a great team of experts in the field who he can learn from and exchange knowledge with. “Imagine that you are building the solution, and after you build it you are going to see the solution work. It is amazing,” he says.
The Mitacs Accelerate program has given Semah the opportunity to work with experts in the field and to engage in a project that he knows will help people and society through its practical application. These two Mitacs internships have allowed Semah to see his CV expand beyond “undergraduate” or “graduate student” to that of machine learning expert or even data analyst.
After being accepted into the Globalink program and progressing to the Accelerate program, Semah is eager to share his experience to motivate other international students to access the many benefits he has gained through working with Mitacs.
For Semah, these benefits include the integration of academic knowledge, access to advanced equipment, learning from the expertise of his professors, and the capacity to improve his programming and his communication skills. Semah has had to learn how to communicate big ideas and concepts to the general public. As he says, “It’s not always easy to explain something that doesn’t yet exist.”
Benefits of applied research for industry and faculty
For Fluent.ai, the internship provides research expertise to help the company further define problems and explore innovative solution strategies.
According to Vikrant Tomar, Founder and CTO of Fluent.ai, partnering with university researchers has multifold advantages.
“First of all,” he says, “you benefit from the enormous research experience of the professors. They bring fresh solutions and technologies that you might not think of yourself. Secondly, it allows you to explore longer-term projects that otherwise would not have been possible in the confined and rather focused research environment in industries.”
He also notes that “Eventually, the interns could return as already trained permanent employees.”
The partnership is so valuable that Fluent.ai has committed to funding many additional internships for this project. And Mitacs is looking for research talent to fill these roles.
For Professor Saucier, the internship programs allow primary investigators to recognize problems that industry partners experience, which can then be a source of inspiration and application for further research.
“The students are a kind of human bridge between the university and the industry, which makes communication both easier and more effective,” says Professor Saucier.
After Semah completes the coding phase of this project and the goal of significant reduction in reverberation in the sound signals is attained, the work will then move to the practical application — the Automatic Speech Recognition system.
As Semah’s work with the Accelerate internship finishes, his coding will continue on in the application of this enhanced system.
Semah, along with his research team from Fluent.ai, Polytechnique Montréal, and Mitacs, hopes this will resonate with you as you access offline voice-activated technology in the future.
Mitacs thanks the Government of Canada for their support of the Accelerate research fellowship in this story. Across Canada, the Accelerate program also receives support from the Government of British Columbia, the Government of Manitoba, the Government of New Brunswick, the Government of Newfoundland and Labrador, the Government of Nova Scotia, the Government of Ontario, the Government of Prince Edward Island, the Government of Quebec, and the Government of Saskatchewan.
Do you have a business challenge that could benefit from a research solution? If so, contact Mitacs today to discuss partnership opportunities: BD@mitacs.ca