Gradient symbolic representations in Harmonic Grammar
Brian Hsu · Language and Linguistics Compass · 2022
Abstract This paper presents an overview of Harmonic Grammar with gradient symbolic representations (a.k.a. Gradient Harmonic Grammar), a weighted‐constraint model of grammatical computation in which language structures are mentally represented with numerically continuous levels of activity, or degree of presence. In this system, the penalty of each constraint violation is proportional to the activity of the structure that incurs it. The adoption of gradient activity permits unique advances in generative approaches to key types of idiosyncratic patterns in language. I review the main proposals that have been made in the framework, and outstanding issues such as the existence of gradient activity in output structures.