Understandable text through machine learning
Funka will explore how machine learning can make it easier to produce texts in plain language. The project is an exciting collaboration with the government owned research institute RISE and is funded by the Swedish innovation agency Vinnova.
It is a major challenge for public sector bodies to produce information that is understandable to everyone. Surveys from the OECD show that around 25% of the adult population in many western countries have difficulties reading and understanding information from authorities and municipalities. This makes it more difficult for many people to participate in society on equal terms. In addition, it affects the authorities' ability to offer services that benefit everyone.
To reach a wider audience, public organisations offer simplified texts in plain language. But producing text that is easier to understand is a specific competence that far from all web authors have. The thousands of texts that authorities and municipalities produce on an daily basis would be very resource demanding to let Funka or other consultancies re-write to meet the requirements.
This is why we want to explore the possibilities of using machine learning to make texts easier to understand.
With the help of new technology, we can enable more people to publish accessible content in their day job as web authors, says Johan Kling, Quality Manager at Funka. It will be exciting to explore the possibilities in this groundbreaking area.
Collaboration with Swedish cutting-edge knowledge in machine learning
Together with leading experts in language technology at the Swedish research institute RISE, we will be training an AI model to develop better texts. The training is done by feeding the model with examples of complicated texts and simplifications of the same texts.
The aim is for the model to support web authors to write simplified texts that avoid complicated wording. The project is the first of its kind in Sweden and will build upon the advanced Swedish language models that RISE is currently developing.
Automatic text simplification is an interesting application of the language models we develop, which we believe can have significant effects to contribute to increased accessibility, says Magnus Sahlgren, Head of Natural Language Processing at RISE.
Funka is looking forward to the collaboration and many new insights.
Period: August 2020 - March 2021
Consortium: Funka (project manager), RISE
Budget: EURO 49 000
Johan KlingTitle: Chief ooperating officer and head of Quality
email@example.com (Johan Kling)