Minority languages such as Welsh lack the extensive electronic resources of major languages. This limits their accessibility, especially for visually impaired people or those with specific communication needs.

The Language Technologies Unit team at Bangor University set out to develop new technologies that would make Welsh more accessible to more people, and at the same time help to revitalise the language.

Developing online resources

The research team developed online resources in two main areas: 

  • Lexica (vocabulary) and grammar
  • Text-to-speech and speech synthesis.

These resources include dictionary apps and proofing, text-to-speech, speech recognition and machine translation tools, such as:

  • Cysgliad, a software package that helps people to write Welsh. It includes a spelling and grammar checker, Cysill, and a set of bilingual dictionaries, Cysgeir.
  • Two Welsh, Windows-based synthetic voices for end-users and Welsh text-to-speech resources that other developers can use to build new voices.
  • Paldaruo, an app to crowdsource speech recordings using smartphones.

The team also worked in partnership with Mozilla on its multilingual Common Voice project. They crowdsourced more than 1,285 Welsh voices, which were then used to develop Macsen, the first Welsh language digital personal assistant, and Trawsgrifiwr, the first Welsh transcriber. 

The team also created the Welsh National Language Technologies Portal to enable people to easily access these resources.

The team’s research has had an impact in four key areas: 

Influencing Policy

Members of the team advised:

  • the Welsh Language Technology Board, on how technology can be used to help revitalise language as part of the Welsh Government’s strategy Cymraeg 2050.
  • the European Parliament’s resolution on language equality in the digital age, which led to a call for digital language equality in Europe by 2030.

Fostering the daily use of Welsh

  • Language technology tools developed by the team are in daily use across Wales. The Welsh Language Commissioner credits Cysgliad with helping people gain confidence in writing Welsh, and as an essential aid in education.

Developing assistive technology

  • The National Library of Wales uses the Trawsgrifiwr programme to help their volunteers transcribe audio material.
  • Organisations such as the Royal National Institute of Blind People (RNIB) use the team’s synthetic Welsh-speaking voices to give visually impaired people access to Welsh text-to-speech.
  • The team’s speech recognition research led to a joint project with the NHS that allows people who are losing their ability to speak to have their voice rebuilt synthetically.

Supporting Industry

  • The LTU team published tools and resources for Welsh speech technology under open licences, which helped commercial companies develop synthetic Welsh-speaking voices.
  • The team took part in a knowledge transfer partnership with a local translation company, Cymen.
  • The team received funding during the COVID-19 pandemic to work with industry partners to develop multilingual translation functions for inclusive online conferencing.

Research team

Professor Delyth Prys, Dewi Bryn Jones, Gruffudd Prys, Dr Sarah Cooper and Dr Myfyr Prys – Bangor University’s Language Technologies Unit (LTU)

Research partners

S4C

Read the full REF impact case study