Papers by Prospero C. Naval, Jr.
Beyond Canonical Fine-tuning: Leveraging Hybrid Multi-Layer Pooled Representations of BERT for Automated Essay Scoring (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing work on automated essay scoring focuses on capturing deep semantic features but are limited to lower-level textual features. |
| Approach: | They propose to use BERT's multi-layer architecture to leverage hierarchical linguistic information from its intermediate layers to improve overall essay scoring performance. |
| Outcome: | The proposed model outperforms the standard model with the default output on the ASAP AES dataset. |