Annotation Scheme for Named Entity Recognition and Relation Extraction Tasks in the Domain of People with Dementia

The number of people suffering from dementia increases with the increase of the aging population. This, in turn, is associated with the need of innovative methods for supporting the family carers of people with dementia (PwD) as the burden of taking care of their relatives results in increased stress and health conditions. Technological solutions could potentially help caregiver to cope with stressful situations caused by the behavior of PwD. In particular, we envision an ontology-based chatbot system that provides recommendations to family carers for coping with the challenging behavior of PwD. The ontology is developed in a semi-automated manner, through manual modeling of domain knowledge and through automated knowledge extraction from textual sources. The challenge in the automated ontology learning is the lack of annotated language resources for this domain. Therefore, in this work we propose an annotation scheme for the extraction of entities and relationships for the domain of agitation and dyadic relationship as a part of stability of care arrangements for PwD. We use the annotation scheme to annotate a textual corpus of informal texts collected from forums for family carers of PwD and we calculate the inter-rater reliability. The results show a moderate overlapping between the annotators. The proposed annotation scheme could serve as a template for providing ground truth for textual datasets in the scarcely annotated domain of PwD.