The automated scoring of narrative essays written by students according to different competences can assist teachers in their evaluation process and help them to focus on specific areas of writing that require improvement among their students. In this paper, we explore the fine-tuning of Portuguese foundation models to automatically score student essays according to four competences: formal register, thematic coherence, narrative rhetorical structure, and cohesion. The results of our experiments show that the agreement between these models and human graders varies between fair and substantial. Thus, although they can provide cues for essay scoring, significant research is still required towards their improvement, especially for the more complex competences.

--

This talk is associated with the following record:

• Exploring the Automated Scoring of Narrative Essays in Brazilian Portuguese using Transformer Models