bert-large-uncased-triage

This model is a fine-tuned version of google-bert/bert-large-uncased for a 5-class triage classification task. It helps categorize student messages based on how they address specific learning objectives.

Classification Labels

0. ADDR_DIRECT: Message directly addresses the objective.
1. ADDR_PARTIAL: Message partially addresses the objective.
2. NOADDR_OFF: Message does not address the objective (Off-topic).
3. NOADDR_ON: Message does not address the objective (On-topic but irrelevant).
4. NOADDR_TANGENTIAL: Message is tangentially related.

Hyperparameters

{
    "learning_rate": 9.424049824827477e-05,
    "num_train_epochs": 3,
    "seed": 7,
    "per_device_train_batch_size": 32
}

Evaluation Results

The model was optimized for Macro-F1 Score on the test set to ensure balanced performance across unique objectives.

Classification Report (test set)

                   precision    recall  f1-score   support

      ADDR_DIRECT      0.875     0.729     0.795        96
     ADDR_PARTIAL      0.571     0.978     0.721        91
       NOADDR_OFF      1.000     0.634     0.776        82
        NOADDR_ON      0.987     0.833     0.904        90
NOADDR_TANGENTIAL      0.873     0.821     0.847        84

         accuracy                          0.801       443
        macro avg      0.861     0.799     0.808       443
     weighted avg      0.858     0.801     0.808       443

Confusion Matrix (test set)