DeepScribe

Habitat Learn Limited (HLL) provides a personalised learning toolkit (HabitatLearn) to improve learning outcomes for students. HabitatLearn is an ecosystem of live closed-captioning, note-taking and mindmaps, enabling students to reach individual learning objectives. It has 20,000+ UK student-users, and 300 HEI partners in UK and North America. DeepScribe builds upon IUK project DeepSpark, an AI toolkit optimising automated speech recognition (ASR) models for higher education.

98% of UK/North American universities were forced online in 2021 and

online/blended learning will remain a permanent feature. Curriculum digitalisation means increasing numbers of international students will attend courses online. Not all students find the transition to online learning easy, especially those with specific learning challenges.

European Accessibility Act (2019) requires accurate video-captioning by public sector bodies to make lectures accessible for students with specific learning disabilities. Despite advances in ASR, human intervention is still required for accuracy, making costs prohibitive.

Workplace and consumer interaction with healthcare advisers and doctors have also been digitalised and individuals with hearing loss suffer the same issues. We will explore this market, piloting DeepScribe with trusted partners.

DeepScribe will develop end-to-end, continuous Personalised Automatic Speech Understanding (PASU) services for private networks and end-user devices using its growing 250,000+ hour repository of human-curated lecture recordings and transcripts across disciplines. Initially targeting educational lecture contexts, but subsequently for wider non-education markets such as healthcare and professional services, where domain knowledge and training data can be covered by our existing medical and health science lectures.

DeepScribe improves current state-of-the-art through:

* Design and implementation of a multi-dimensional context-aware personalised ASR training process for individual users;
* Optimising PASU fine-tuning algorithms using low-specification computing instead of high-end GPUs and TPUs;
* Using edge computing architecture to maximise security and privacy for audio data collection and PASU training , limiting personal data-sharing on the cloud;

* accurate transcription to enable mass-affordability for different use cases.

DeepScribe will provide personalised automation of speech recognition applied to education, the workplace and everyday interaction with healthcare advisers/doctors to improve accuracy, disciplinary specificity, and speed of captioning.

DeepScribe will use incremental AI-based fine-tuning and edge computing to deliver PASU with speed of response and affordability, resulting in a commercially viable proposition.

DeepScribe will support mass-adoption of HLL's technology in education, transforming it from a limited accessibility tool to a generic learning toolkit.

DeepScribe will broaden HLL's customer base outside of education.

HLL will create 15 highly skilled UK jobs over 5 years.