A Comparative Study of Modal Verb Frameworks with Annotated Datasets

Anonymous

Abstract: Modal verbs, such as $\textit{can}$, $\textit{may}$, and $\textit{must}$, are commonly used in our daily communication to convey the speaker's perspective related to the likelihood and/or mode of the proposition. They can differ greatly in meaning depending on how they're used and the context of a sentence (e.g. ``They $\textit{must}$ help each other out.'' vs. ``They $\textit{must}$ have helped each other out.'')Despite their practical importance in areas such as natural language understanding, linguists have yet to agree on a single, prominent framework for the categorization of modal verb senses. This lack of agreement stems from high degrees of flexibility and polysemy from the modal verbs, making it more difficult for researchers to incorporate insights from this family of words into their work.As a tool to help navigate this issue, we present MoVerb, which consists of 4.5K annotated sentences from social conversations in Empathetic Dialogues, with each sentence being annotated using two different theoretical frameworks of modal verb senses. We offer insight into the challenges of modal verb ambiguity and suggest modifications when annotating them for downstream NLP tasks.Our dataset will be publicly available upon acceptance.

Paper Type: long