mix nasty.train.crf (Nasty v0.3.0)

Trains a CRF (Conditional Random Field) model for sequence labeling tasks.

Usage

mix nasty.train.crf --corpus data/train.conllu --output priv/models/en/ner_crf.model --task ner

Options

• --corpus - Path to training corpus in CoNLL-U format (required)
• --test - Path to test corpus for evaluation (optional)
• --output - Path to save trained model (required)
• --task - Task type: ner, pos, chunking (default: ner)
• --iterations - Maximum training iterations (default: 100)
• --learning-rate - Learning rate (default: 0.1)
• --regularization - L2 regularization strength (default: 1.0)
• --method - Optimization method: sgd, momentum, adagrad (default: momentum)
• --language - Language code (default: en)

Examples

# Train NER model
mix nasty.train.crf \
  --corpus data/en_ewt-ud-train.conllu \
  --output priv/models/en/ner_crf.model \
  --task ner \
  --iterations 100

# Train with evaluation
mix nasty.train.crf \
  --corpus data/train.conllu \
  --test data/test.conllu \
  --output priv/models/en/ner_crf.model \
  --task ner \
  --learning-rate 0.05