ImpScore: A Learnable Metric For Quantifying The Implicitness Level of Sentence

Understand

Handling implicit language is essential for natural language processing systems to achieve precise text understanding and facilitate natural interactions with users.

• Despite its importance, the absence of a metric for accurately measuring the implicitness of language significantly constrains the depth of analysis possible in evaluating models' comprehension capabilities.
• This paper addresses this gap by developing a scalar metric that quantifies the implicitness level of language without relying on external references.
• Drawing on principles from traditional linguistics, we define "implicitness" as the divergence between semantic meaning and pragmatic interpretation.