New innovators: SoapBox Labs platform personalises child reading lessons

Speech recognition tech ‘listens’ to assess reading level, lessons can then be adjusted

Patricia Scanlon, founder of SoapBox Labs: collected voice samples from 9,000 children to underpin her technology
Patricia Scanlon, founder of SoapBox Labs: collected voice samples from 9,000 children to underpin her technology

Learning to read is a huge step in every child's development and most children get there sooner or later. But Patricia Scanlon, founder of SoapBox Labs, believes speech recognition technology has a big part to play in making this milestone easier to achieve.

Scanlon holds a PhD in speech technologies and her area of specialisation is speech recognition. While teaching her daughter to read, Scanlon discovered that the apps and web services available to help build children’s reading and language skills were of the “one size fits all” variety and did not take individual learning needs into account.

She also found that the speech recognition products already on the market had a number of shortcomings. Specifically, they could only be used for children who had mastered reading full words. (This meant those in the 4- 8-year-old age group were not being well catered for.) The systems were also highly sensitive to background noise and were less accurate as a result. Finally, most were not designed for mobile technology and had to be used with headsets.

This combination of deficits led Scanlon to start thinking about a better solution. She began gathering the voice samples that would underpin her technology and spent more than two years building a data bank of half a million samples from 9,000 children in 89 countries. With this complete data set she started putting the other pieces in place and is now in the final stages of developing a cloud-based proprietary children’s speech recognition platform that addresses all of the problems she had encountered with existing systems.

READ MORE

The SoapBox platform automatically assesses the child’s reading skills by “listening” to each child and analysing their pronunciation. Based on the results, their reading and language lessons can be adjusted in real time to ensure that each child is working with the right degree of difficulty for them.

SoapBox Labs is unusual in that it is a spin-in to Trinity College Dublin. "I needed highly qualified people to work with me on the technology and was keen to work closely with the Learnovate Technology Centre so in late 2014 I became the college's first spin-in entrepreneur," Scanlon says.

"By coming here we have been able to develop our solution in an environment specifically designed to promote innovation in learning technology. We officially became a Trinity campus company in December 2015 and now have four people on the team including myself - speech scientist Amelia Kelly, technology lead Robert O'Regan and software developer Ronan Tumelty."

Scanlon is no stranger to making the jump from research bench to market having most recently worked with Bell Labs where her role involved working with research engineers, business groups and university collaborators to commercialise research innovations.

Scanlon’s route to market is through education publishers who will license the technology for use with mobile apps and other web-based products. The product will be launched in September 2016 and Scanlon estimates the value of the global early literacy and language market at about $1.5 billion.

SoapBox has been supported by Enterprise Ireland’s Commercialisation Fund and development costs to date are more than €360,000. The company in now looking for about €750,000 in outside investment. Heads of agreement have been drawn up with its first Irish customer while discussions are at an advanced stage with UK and US publishers.

Competition comes from printed materials and from existing apps and web services. However, they don’t have the automatic assessment tool or the personalised learning paths. There is also competition from existing off-the-shelf speech recognition products but unlike SoapBox, they can’t recognise phonics, blended sounds or syllables.

“Our platform provides robust performance in real world noise environments such as classrooms and homes and it has been designed to work on mobile phones and tablets without headsets/microphones which is how most children interact with technology today,” Scanlon says.

– OLIVE KEOGH