How Curriculum Learning Performs on AMR Parsing

Abstract

Curriculum learning is a commonly used method in deep learning to improve training model efficiency. This method has been proven effective on a wide range of tasks in natural language and image processing. However, there are no studies yet fully investigating the possibility of applying this method to AMR parsing, the task of converting a sentence into an AMR, its abstract meaning representation. In this article, we experiment and investigate in detail the possibilities of applying curriculum into AMR parsing.