Linear training

usage: kiwi train [-h] [--train-source TRAIN_SOURCE]
                  [--train-target TRAIN_TARGET]
                  [--train-alignments TRAIN_ALIGNMENTS]
                  [--train-source-tags TRAIN_SOURCE_TAGS]
                  [--train-target-tags TRAIN_TARGET_TAGS]
                  [--train-source-pos TRAIN_SOURCE_POS]
                  [--train-target-pos TRAIN_TARGET_POS]
                  [--train-target-parse TRAIN_TARGET_PARSE]
                  [--train-target-ngram TRAIN_TARGET_NGRAM]
                  [--train-target-stacked TRAIN_TARGET_STACKED]
                  [--valid-source VALID_SOURCE] [--valid-target VALID_TARGET]
                  [--valid-alignments VALID_ALIGNMENTS]
                  [--valid-source-tags VALID_SOURCE_TAGS]
                  [--valid-target-tags VALID_TARGET_TAGS]
                  [--valid-source-pos VALID_SOURCE_POS]
                  [--valid-target-pos VALID_TARGET_POS]
                  [--valid-target-parse VALID_TARGET_PARSE]
                  [--valid-target-ngram VALID_TARGET_NGRAM]
                  [--valid-target-stacked VALID_TARGET_STACKED]
                  [--source-vocab-size SOURCE_VOCAB_SIZE]
                  [--target-vocab-size TARGET_VOCAB_SIZE]
                  [--source-vocab-min-frequency SOURCE_VOCAB_MIN_FREQUENCY]
                  [--target-vocab-min-frequency TARGET_VOCAB_MIN_FREQUENCY]
                  [--use-basic-features-only USE_BASIC_FEATURES_ONLY]
                  [--use-bigrams USE_BIGRAMS]
                  [--use-simple-bigram-features USE_SIMPLE_BIGRAM_FEATURES]
                  [--training-algorithm TRAINING_ALGORITHM]
                  [--regularization-constant REGULARIZATION_CONSTANT]
                  [--cost-false-positives COST_FALSE_POSITIVES]
                  [--cost-false-negatives COST_FALSE_NEGATIVES]
                  [--evaluation-metric EVALUATION_METRIC]

data

--train-source	Path to training source file
--train-target	Path to training target file
--train-alignments
	Path to train alignments between source and target.
--train-source-tags
	Path to validation label file for source (WMT18 format)
--train-target-tags
	Path to validation label file for target
--train-source-pos
	Path to training PoS tags file for source
--train-target-pos
	Path to training PoS tags file for target
--train-target-parse
	Path to training dependency parsing file for target (tabular format)
--train-target-ngram
	Path to training highest order ngram file for target (tabular format)
--train-target-stacked
	Path to training stacked predictions file for target (tabular format)

validation data

--valid-source	Path to validation source file
--valid-target	Path to validation target file
--valid-alignments
	Path to valid alignments between source and target.
--valid-source-tags
	Path to validation label file for source (WMT18 format)
--valid-target-tags
	Path to validation label file for target
--valid-source-pos
	Path to training PoS tags file for source
--valid-target-pos
	Path to training PoS tags file for target
--valid-target-parse
	Path to validation dependency parsing file for target (tabular format)
--valid-target-ngram
	Path to validation highest order ngram file for target (tabular format)
--valid-target-stacked
	Path to validation stacked predictions file for target (tabular format)

vocabulary options

--source-vocab-size
	Size of the source vocabulary.
--target-vocab-size
	Size of the target vocabulary.
--source-vocab-min-frequency
	Min word frequency for source vocabulary. Default: 1
--target-vocab-min-frequency
	Min word frequency for target vocabulary. Default: 1

linear

Linear Quality Estimation

--use-basic-features-only
	1 for using only basic features (words). Default: 0
--use-bigrams	1 for using bigram features (i.e. a CRF-like model). Default: 1
--use-simple-bigram-features
	1 for using only label indicators as bigram features. Default: 0
--training-algorithm
	Algorithm for training the model (svm_mira, svm_sgd, perceptron). Default: “svm_mira”
--regularization-constant
	L2 regularization constant. Default: 0.001
--cost-false-positives
	Cost for false positives (svm_mira and svm_sgd only). Default: 0.2
--cost-false-negatives
	Cost for false negatives (svm_mira and svm_sgd only). Default: 0.8
--evaluation-metric
	Evaluation metric (f1_mult or f1_bad). Default: “f1_mult”