mix nasty.eval.pos (Nasty v0.3.0)

View Source

Evaluates a trained POS tagging model on test data.

Usage

mix nasty.eval.pos --model MODEL_PATH --test TEST_FILE [options]

Options

--model PATH        Path to trained model file (required)
--test PATH         Path to CoNLL-U test file (required)
--baseline          Also evaluate rule-based baseline for comparison

Examples

# Evaluate a trained model
mix nasty.eval.pos \
  --model priv/models/en/pos_hmm_v1.model \
  --test data/UD_English-EWT/en_ewt-ud-test.conllu

# Compare with rule-based baseline
mix nasty.eval.pos \
  --model priv/models/en/pos_hmm_v1.model \
  --test data/UD_English-EWT/en_ewt-ud-test.conllu \
  --baseline

Output

The task reports:

  • Overall accuracy
  • Macro-averaged F1, precision, and recall
  • Per-class performance for each POS tag
  • Top and bottom performing tags
  • Confusion matrix (optional)

If --baseline is provided, compares the model against rule-based tagging.