Evaluator¶
-
class
ucca.evaluation.
Evaluator
(verbose, constructions, units, fscore, errors)[source]¶ Bases:
object
Methods Summary
find_mutuals
(m1, m2, eval_type, mutual_tags)get_scores
(p1, p2, eval_type[, r])prints the relevant statistics and f-scores. Methods Documentation
-
get_scores
(p1, p2, eval_type, r=None)[source]¶ prints the relevant statistics and f-scores. eval_type can be ‘unlabeled’, ‘labeled’ or ‘weak_labeled’. calculates a set of all the yields such that both passages have a unit with that yield. :param p1: passage to compare :param p2: reference passage object :param eval_type: evaluation type to use, out of EVAL_TYPES 1. UNLABELED: it doesn’t matter what labels are there. 2. LABELED: also requires tag match (if there are multiple units with the same yield, requires one match) 3. WEAK_LABELED: also requires weak tag match (if there are multiple units with the same yield,
requires one match)Parameters: r – reference passage for fine-grained evaluation Returns: EvaluatorResults object if self.fscore is True, otherwise None
-