Comparative Analysis of Zero-Shot and Task-Specific Transformer Models in Argumentation Analysis

• In Event: GSIC Poster Session II

Abstract

In this paper, we present a comparison between a zero-shot language model and a task-specific transformer model specifically trained on the DARIUS corpus. Both models predict argumentative structures in German high school essays with the goal of providing adaptive feedback. We measure their performance using the same DARIUS testing corpus.

Author

• Nils-Jonathan Schaller, IPN – Leibniz Institute for Science and Mathematics Education