We have been successfully developing a model to //measure// the readability of Wikipedia articles ([[ https://meta.wikimedia.org/wiki/Research:Multilingual_Readability_Research | project-page ]] on metawiki). As a next step, we would like to develop a model that could improve the readability of Wikipedia articles (along the lines of text simplification) taking advantage of recent advances in availability and performance of large language models.
As a first step, in this task, we want to scope the project in more detail. Specifically, we would like to get a better overview of potential approaches for implementation.
[x] Reviewing recent literature
[x] Reviewing existing models for text summarization and text simplification and comparing with available infrastructure on, e.g., LiftWing, to train/host model
[x] Reviewing approaches for evaluation
[x] Identify relevant benchmark/evaluation datasets
[ ] From the above, synthetize a work-plan for implementing and testing an exploratory model for simplification