Taaco-BERT: A Linguistically Interpretable Large Language Model for Essay Scoring in Educational Settings

Poster Session VII - Research Proposal Posters

Keyword: Literacy

Abstract: We propose TAACO-BERT, integrating TAACO language measures with continuous pre-training and fine-tuning of BERT-based models on three educational datasets. This approach improves the models’ ability to capture linguistic features and knowledge construction, enhancing predictions for writing-based educational assessments.

• Heqiao Wang (Presenting Author)
• Kevin Haudek (Author)
• Shirley Vincent (Author)
• Caterina Azzarello (Author)
• Steven Anderson (Author)