Cost-Efficient Subjective Task Annotation and Modeling through Few-Shot Annotator Adaptation

arXiv:2402.14101v1 Announce Type: new
Abstract: In subjective NLP tasks, where a single ground truth does not exist, the inclusion of diverse annotators becomes crucial as their unique perspectives significantly influence the annotations. In realistic scenarios, the annotation budget often becomes the main determinant of the number of perspectives (i.e., annotators) included in the data and subsequent modeling. We introduce a novel framework for annotation collection and modeling in subjective tasks that aims to minimize the annotation budget while maximizing the …

abstract annotation annotations arxiv budget cost cs.cl diverse few-shot inclusion influence modeling nlp perspectives tasks through truth type