README.md

metadata

metrics:
  - accuracy
base_model:
  - aubmindlab/bert-base-arabertv2
pipeline_tag: token-classification

AraBERT-based Discourse Connective Classifier

This model detects whether a given Arabic token is a discourse connective based on AraBERT embeddings and additional handcrafted features (Is_potential_conn, Has_Almasdar).

Model Details

• Base model: aubmindlab/bert-base-arabertv02
• Architecture: BERT embeddings + 2-dimensional feature projector + classification head
• Input: List of tokenized words for each sentence, plus features
• Output: Yes (connective) or No (not connective) for each token