Avanade is participating in the Innovate UK-funded DeepMyna project, a groundbreaking initiative aimed at developing more trustworthy AI tools by identifying and mitigating bias in speech recognition systems.
The project, led by Habitat Learn, has been awarded a £1.6 million grant from Innovate UK and includes contributions from Avanade, Affiniti AI, Microlink PC and the University of Southampton.
The research aims to develop solutions that increase trust in AI, with a focus on improving Automatic Speech Recognition (ASR) systems by enhancing accuracy and reducing bias. With applications across industries such as healthcare, education, finance and legal services, the initiative has the potential to transform how businesses and individuals use AI-powered tools.
The grant is being utilised by the consortium members to fund work in their respective areas of expertise. Habitat Learn is spearheading the technical development, Avanade is sourcing data for testing and supporting the product and market strategy, Affiniti AI is contributing healthcare-related data, Microlink PC is offering assistive technology expertise and the University of Southampton is providing renowned researchers in the speech recognition space.
Through their combined capabilities, the consortium is working to ensure the commercialisation of research outputs, ultimately delivering significant economic benefits for the UK while fostering greater trust in AI.
Martin Harper, UKI Innovation Lead at Avanade, said: “The DeepMyna project represents a significant step forward in addressing bias in speech recognition systems and we are proud to contribute our expertise in data sourcing and strategic analysis as part of this consortium. Together, we aim to develop AI tools that all users can trust and rely on with confidence.”
Jeremy Brassington, CEO, Habitat Learn Limited and Consortium Lead for the project said: “Avanade’s involvement brings valuable expertise in data sourcing and strategic analysis, essential for driving the success of this initiative. This project aligns with our mission to drive productivity and economic growth by supporting businesses in developing innovative and trustworthy AI solutions.”
The DeepMyna project is expected to be completed by the end of January 2025. Once the research output has been fully evaluated, the consortium plans to apply their learnings and develop use cases that deliver the most value. The research focuses on enhancing the accuracy of AI speech recognition tools by reducing transcription Word Error Rate and increasing trust in AI systems. This is achieved by identifying biases and providing recommendations for fine-tuning. It empowers users of speech-to-text technology to improve the accuracy and inclusivity of their tools by intentionally modifying the training data used.
The consortium has identified several use cases for the research outputs, including:
- Communications and Media: Creating intelligent consumer product experiences.
- Cross-Industry Applications: Refining customer service interactions, virtual assistants, and transcription tools.
- Education: Improving captioning, transcription, and interactive learning experiences.
- Finance: Advancing voice biometrics for fraud detection.
- Healthcare: Enhancing patient interactions, note-taking, documentation, and assistive technologies.
- Industrials: Improving accuracy of technical transcriptions to support international engineering collaborations.
- Legal Services: Supporting live deposition and court transcription accuracy.
