Two Contrasting Data Annotation Paradigms for Subjective NLP Tasks
This addresses the issue of partly-subjective datasets failing to serve clear downstream uses for NLP researchers and dataset creators, though it is incremental in proposing explicit paradigms rather than a new method.
The paper tackles the problem of annotator subjectivity in NLP tasks by proposing two contrasting data annotation paradigms: descriptive, which encourages subjectivity to model diverse beliefs, and prescriptive, which discourages it to train consistent models. An annotation experiment with hate speech data illustrates the contrast between these paradigms.
Labelled data is the foundation of most natural language processing tasks. However, labelling data is difficult and there often are diverse valid beliefs about what the correct data labels should be. So far, dataset creators have acknowledged annotator subjectivity, but rarely actively managed it in the annotation process. This has led to partly-subjective datasets that fail to serve a clear downstream use. To address this issue, we propose two contrasting paradigms for data annotation. The descriptive paradigm encourages annotator subjectivity, whereas the prescriptive paradigm discourages it. Descriptive annotation allows for the surveying and modelling of different beliefs, whereas prescriptive annotation enables the training of models that consistently apply one belief. We discuss benefits and challenges in implementing both paradigms, and argue that dataset creators should explicitly aim for one or the other to facilitate the intended use of their dataset. Lastly, we conduct an annotation experiment using hate speech data that illustrates the contrast between the two paradigms.