Plain language support through machine learning
| Reference number | |
| Coordinator | Funka nu AB |
| Funding from Vinnova | SEK 499 340 |
| Project duration | August 2020 - March 2021 |
| Status | Completed |
| Venture | AI - Competence, ability and application |
| Call | Start your AI-journey! For businesses |
Important results from the project
The purpose of the project was to develop a method for making Swedish texts from public organizations more accessible through simplification with the help of machine learning. In the project, we collected and organised a large number of documents including complicated and simplified versions of the same text. These examples were used to train a model for machine learning in simplifying texts. The objective was to test whether machine learning can provide support for communicators working with plain language.
Expected long term effects
The expected effects in the long run were that it would be easier and cheaper for public organizations to produce texts in plain language with the help of a tool for automatic processing of texts. But in the project, we discovered that such a tool could not be developed based on the data we had collected in combination with the existing methods and language models that are available for the Swedish language. On the other hand, the project has given a boost to the company´s knowledge of AI. It has also made a contribution to scientific research on models for language simplification in Swedish.
Approach and implementation
The collection of data is a large part of the proceeds of the project, both because we now have a large and well-organized data base and because the collection has created commitment among actors in plain language. The next step was to train the machine learning models with the collected data. However, result of the training showed that the structure for plain language processing is not compatible with the models we had available in the project. We have therefore concluded that more experiments are needed in the field to find out how machine learning can be applied to plain language.