Parser

class tupa.parse.Parser(model_files=(), config=None, beam=1)

Bases: tupa.parse.AbstractParser

Main class to implement transition-based UCCA parser

Methods Summary

eval(passages, mode, scores_filename[, display])
eval_and_save([last, finished_epoch])
init_train()
parse(passages[, mode, evaluate, display, write])	Parse given passages :param passages: iterable of passages to parse :param mode: ParseMode value.
print_config()
save(model)
train([passages, dev, test, iterations])	Train parser on given passages :param passages: iterable of passages to train on :param dev: iterable of passages to tune on :param test: iterable of passages that would be tested on after train finished :param iterations: iterable of Iterations objects whose i attributes are the number of iterations to perform

Methods Documentation

eval(passages, mode, scores_filename, display=True)

eval_and_save(last=False, finished_epoch=False)

init_train()

parse(passages, mode=<ParseMode.test: 3>, evaluate=False, display=True, write=False)

Parse given passages :param passages: iterable of passages to parse :param mode: ParseMode value.

If train, use oracle to train on given passages. Otherwise, just parse with classifier.

Parameters:

• evaluate – whether to evaluate parsed passages with respect to given ones. Only possible when given passages are annotated.
• display – whether to display information on each parsed passage
• write – whether to write output passages to file

Returns:

generator of parsed passages (or in train mode, the original ones), or, if evaluation=True, of pairs of (Passage, Scores).

print_config()

save(model)

train(passages=None, dev=None, test=None, iterations=1)

Train parser on given passages :param passages: iterable of passages to train on :param dev: iterable of passages to tune on :param test: iterable of passages that would be tested on after train finished :param iterations: iterable of Iterations objects whose i attributes are the number of iterations to perform