Recent work has explored the ability of large language models (LLMs) to generate explanations of existing labeled data. In this work, we investigate the ability of LLMs to explain revisions in sentences. We introduce a new dataset demonstrating a novel task, which we call explaining text revisions. We collected human- and LLM-generated explanations of grammatical and fluency edits and defined criteria for the human evaluation of the explanations along three dimensions: Coverage, Informativeness, and Correctness. The results of a side-by-side evaluation show an Overall preference for human explanations, but there are many instances in which annotators show no preference. Annotators prefer human-generated explanations for Informativeness and Correctness, but they show no preference for Coverage. We also examined the extent to which the number of revisions in a sentence influences annotators’ Overall preference for the explanations. We found that the preference for human explanations increases as the number of revisions in the sentence increases. Additionally, we show that the Overall preference for human explanations depends on the type of error being explained. We discuss explanation styles based on a qualitative analysis of 300 explanations. We release our dataset and annotation guidelines to encourage future research.

Annotation of discourse relations is a known difficult task, especially for non-expert annotators. In this paper, we investigate novice annotators’ uncertainty on the annotation of discourse relations on spoken conversational data. We find that dialogue context (single turn, pair of turns within speaker, and pair of turns across speakers) is a significant predictor of confidence scores. We compute distributed representations of discourse relations from co-occurrence statistics that incorporate information about confidence scores and dialogue context. We perform a hierarchical clustering analysis using these representations and show that weighting discourse relation representations with information about confidence and dialogue context coherently models our annotators’ uncertainty about discourse relation labels.

Time pressure and topic negotiation may impose constraints on how people leverage discourse relations (DRs) in spontaneous conversational contexts. In this work, we adapt a system of DRs for written language to spontaneous dialogue using crowdsourced annotations from novice annotators. We then test whether discourse relations are used differently across several types of multi-utterance contexts. We compare the patterns of DR annotation within and across speakers and within and across turns. Ultimately, we find that different discourse contexts produce distinct distributions of discourse relations, with single-turn annotations creating the most uncertainty for annotators. Additionally, we find that the discourse relation annotations are of sufficient quality to predict from embeddings of discourse units.