Skip to main content
European Commission logo
Deutsch Deutsch
CORDIS - Forschungsergebnisse der EU
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary

Multilingual Children’s Speech Assessment Platform for Literacy and Language Learning

Periodic Reporting for period 2 - SpeechTech4Literacy (Multilingual Children’s Speech Assessment Platform for Literacy and Language Learning)

Berichtszeitraum: 2019-03-01 bis 2020-02-29

Reading is a foundational skill. The ability to “read to learn” is the key to unlocking the potential knowledge and skills of children. A child’s proficiency at reading by age 8 is a key indicator of their future success. Up until age 8 children are learning to read; by age 8 they start reading to learn. Researchers at Yale University highlighted the importance of catching children with poor reading skills early, as 75% of poor readers at age 8 will remain poor readers in high school. Studies have shown that low literacy is more likely to cause children to: leave school early; be unemployed; have low-skilled jobs; end up in poverty; and have poorer emotional and physical health.

An OECD study shows that approximately one in five 15-year olds across the EU have difficulties using reading as a tool for learning. Worryingly, this report also highlighted that the EU has made almost no progress over the past decade in reducing this percentage. However, these average statistics mask a significantly worse problem: literacy levels of disadvantaged and immigrant children. The OECD studies have consistently shown that socio-economic disadvantage is linked to poor performance in school, on average a difference of more than two years of schooling. In 2012, 11% of 15-year-old students had an immigrant background, and typically from disadvantaged backgrounds. Reading levels for these immigrant students lag far behind those for non-immigrant students.

Irish company SoapBox Labs (SBL) is addressing this problem by developing a multilingual children’s speech assessment platform specifically for young children. This platform is available for license to ALL creators of educational products e.g. education publishers, Education Technology (EdTech) companies and independent developers to integrate into their digital literacy and language learning products. The speech assessment technology works robustly on consumer mobile devices, in real world noise environments such as homes and schools. Our technology platform, with an open licensing model, has the potential to reach large numbers of children across the EU and beyond in order to have a REAL impact on reading and language learning skills.

Our children’s multilingual speech assessment platform is available in multiple languages, to enable voice interactions and speech (pronunciation) assessment for children’s educational digital products, for all stages of learning: from preliterate to emergent and mature readers.

SBL’s vision: To enact real social change by delivering an AI-driven, scalable and cost effective way to teach reading and language learning to children globally using our speech recognition technology.

Key Project Objectives
- Assessment platform to include preliterate, emergent and mature readers
- Scaling platform in multiple languages
- Validating speech assessment through school based tails
- Commercial readiness for global product deployment
All activities of the SpeechTech4Literacy project have been on track, all of the tasks we were to complete by the end of the project have been fulfilled on time, and all deliverables of the period submitted.
At the beginning of the project we had English only infrastructure and data. By the end of first year, we had adapted adapted and scaled our system to be ‘language agnostic’ - our infrastructure can adapt to any global language. Our API was also adapted to support multiple languages.
By the end of the project, we have completed all the specific objectives we set out. We have collected sufficient amount of data in all languages; our speech assessment platform is operational in all languages and has also been trialed; finally, our market deployment plan is updated and we have a strong pipeline of clients. We have devised a structured approach to our communications strategy in terms of targeting different verticals and markets. This also aligns to our product work of market segmentation, persona creation, business model canvasses and value propositions.
SoapBox Labs is pushing the boundaries of state of the art in speech recognition in two principle ways: firstly, in the use of Deep Neural Networks – applied to our large and proprietary multi-lingual data set - to deliver significant improvements in accuracy for children’s speech technology; and, secondly, in the delivery mechanisms used to deploy the technology for end user benefit (with our cloud-based infrastructure and offline using the same deep learning techniques – both embedded and on device).
The wider societal implications for these innovations can be summarised in terms of ability to scale and cost-effective applications to support multiple use cases for young children in developed and developing countries. These use cases include: assistive technologies for accessibility; early childhood literacy; language learning; and, diagnostic tools to identify reading difficulties (e.g. dyslexia screeners) and developmental delays.
Kids using our speech tech