The paper presents a study on combining model- and data-centric approaches to
building a question answering system for inclusion of people with autism spectrum
disorder. The study shows that applying sequentially model- and data-centric
approaches might allow achieving higher metric scores on closed-domain lowresourced datasets.