Alveo: Above and beyond speech, language and music

Alveo: Above and beyond speech, language and music
June 8, 2016 Karen Mecoles

“It’s important to understand whether people need to change to fit machines, or if machines need to change to fit people”

Professor Denis Burnham

Exploring the relationship between human and mechanised communication, the relatively new research field of human communication science (HCS) brings together phonetics, linguistics, psychology, language technology, computer science, and music cognition to solve key health, social and cultural communication problems.

Following the Australian Research Council’s call for more collaborative, interdisciplinary research projects, the field of HCS took off in Australia in 2003, led by Professor Denis Burnham, Director of the MARCS Institute at the Western Sydney University.

“As a result of the ARCs call for research collaboration, my colleagues and I, who were studying speech and music cognition at MARCS, joined forces with researchers specialising in language technology and computer science at Macquarie University,” Professor Burnham said.

Now President of the Australasian Speech Science & Technology Association (ASSTA), Professor Burnham takes an HCS approach to his research, studying language development in children and differences in the ways people speak to infants, foreigners, pets, lovers, and computers.

“HCS is more or less the way humans communicate with each other and with machines in codified manners: speech, text, music, or emotional responses to sounds,” Professor Burnham said. “Ten to fifteen years ago I wouldn’t have thought I’d be working on things like this,” he said. “It’s certainly been a learning curve, for all of us.

“By bringing researchers from each of these fields together you get a confluence of ideas and find new ways of approaching problems – it leads to serendipity.”

An increasingly important field, HCS enables researchers to program machines to better understand what people are saying, doing and trying to communicate. This includes everything from automated speech recognition and cochlear implants, to learning aids for disabled children.

“When you’re trying to fit machines and people together it’s important to understand whether people need to change to fit machines, or if machines need to change to fit people,” Professor Burnham said.

In response to the ARCs call for collaborative Research Networks, Professor Burnham and his colleagues formed HCSNet, a five-year ARC funded program to help stimulate the field of HCS in Australia.

“As we were really at the coalface of this new field of research the HCSNet held over 60 seminars and workshops and five annual conferences over the five years it was in operation,” Professor Burnham said.

“As a result of the network, a number of collaborative research projects were established, including AusTalk, an auditory-visual database of Australian English collected from 1000 speakers and funded under the Linkage Infrastructure, Equipment and Facilities (ARC LIEF) program.

“We had effectively formed a community of researchers working on common problems in HCS, but lacked a common infrastructure to work from.”

While significant data collections and analysis tools from projects such as AusTalk were being created and collected by the HCS research community, the disparate locations of the research institutions involved made collaboration and replication difficult.

“That’s when we discovered Nectar Labs,” Professor Burnham said. “A virtual space for us to build and store our collections, add our analysis and workflow tools and ensure accessibility across the county, and even around the world.”

“We launched our virtual laboratory, Alveo, in June 2014, providing all researchers with the infrastructure to store data collections, analysis tools, and workflows in a common environment, allowing all HCS researchers to study speech, language, text, and music on a larger scale.”

Funded by Nectar, the Alveo virtual laboratory was established by Associate Professor Steve Cassidy from Macquarie University (Product Owner) and Dr Dominique Estival from UWS (Project Manager) with the help of Intersect Australia, who undertook the software development.

“With 13 different universities involved in the Alveo partnership, the set-up of the virtual lab was a challenge,” Professor Burnham said.

“At different institutions there were different specialties, Melbourne had a lot of linguist’s and engineers, Macquarie had a number of language and technology researchers, while at UWS we were focusing more on psychology and music cognition.”

The Alveo launch at the end of June included a ‘hackfest’ and user workshop, allowing researchers to explore ways they could use the virtual lab, and come up with ideas for new tools and data to incorporate into the program.

“I was struck by the number and diversity of researchers interested in adding collections and using the virtual lab in their work,” he said. “Alveo has really inspired me to consider new ways of approaching my research.

“If all HCS researchers put their data in the one place, with the same metadata and a suite of analysis tools, we could standardise research so much more easily and have far greater outcomes, it’s really exciting.”

Alveo is now entering its second phase with UWS partnering with the recently established ARC Centre of Excellence for the Dynamics of Language, directed by Professor Nicholas Evans at The Australian National University. Alveo will also become a portal for PARADISEC, the Pacific and Regional Archive for Digital Sources in Endangered Cultures, a key facility for preserving recordings of endangered languages and music.

For more information about Alveo, go to