Development of an Automated, Rule-Based Measurement Method for Easy Language and Its Application to AI-Generated Texts

Easy Language improves information accessibility, especially for those with limited language proficiency or reading skills. Rules for Easy Language are collaboratively developed and undergo comprehension checks by individuals with intellectual disabilities. This paper introduces an automated method to assess Easy Language usage in AI-generated texts. Initial manual rules are translated into automatable criteria. An automated system applies these rules to texts, distinguishing between Easy Language and complex language. Thresholds for this differentiation are established using a development set, incorporating expert input and analyzing various text types. The system is then tested on existing texts, including those translated by experts and AI services like SUMM AI and capito general. Future work aims to expand this approach for inclusive AI, facilitating interaction between individuals with disabilities and technology.