o 5tiP@sdZddlmZddlmZddlmZddlZddlZddlmZddl Z ddl Z ddl m Z Gdd d ed gd ZGd d d eejd ZGdddedgdZGdddeZddZdS)zLibrary for scoring and evaluation of text samples. Aggregation functions use bootstrap resampling to compute confidence intervals as per the original ROUGE perl implementation. )absolute_import)division)print_functionN)Dict)rangec@eZdZdZdS)Scorez9Tuple containing precision, recall, and f-measure values.N__name__ __module__ __qualname____doc__rrG/home/ubuntu/.local/lib/python3.10/site-packages/rouge_score/scoring.pyr"r) precisionrecallfmeasurec @s4eZdZdZejdededeeeffddZ dS) BaseScorerzBase class for Scorer objects.target predictionreturncCsdS)aCalculates score between the target and prediction. Args: target: Text containing the target (ground truth) text. prediction: Text containing the predicted text. Returns: A dict mapping each score_type (string) to Score object. Nr)selfrrrrrscore*szBaseScorer.scoreN) r r r r abcabstractmethodstrrrrrrrrr's$r) metaclassc@r)AggregateScorez1Tuple containing confidence intervals for scores.Nr rrrrr7rrlowmidhighc@s2eZdZdZd ddZddZdd Zd d Zd S)BootstrapAggregatoraAggregates scores to provide confidence intervals. Sample usage: scorer = rouge_scorer.RougeScorer(['rouge1', 'rougeL']) aggregator = Aggregator() aggregator.add_scores(scorer.score("one two three", "one two")) aggregator.add_scores(scorer.score("one two five six", "seven eight")) result = aggregator.aggregate() print result {'rougeL': AggregateScore( low=Score(precision=0.0, recall=0.0, fmeasure=0.0), mid=Score(precision=0.5, recall=0.33, fmeasure=0.40), high=Score(precision=1.0, recall=0.66, fmeasure=0.80)), 'rouge1': AggregateScore( low=Score(precision=0.0, recall=0.0, fmeasure=0.0), mid=Score(precision=0.5, recall=0.33, fmeasure=0.40), high=Score(precision=1.0, recall=0.66, fmeasure=0.80))} ffffff?cCsD|dks|dkr td|dkrtd||_||_tt|_dS)aInitializes a BootstrapAggregator object. Args: confidence_interval: Confidence interval to compute on the mean as a decimal. n_samples: Number of samples to use for bootstrap resampling. Raises: ValueError: If invalid argument is given. rz+confidence_interval must be in range [0, 1]zn_samples must be positiveN) ValueError _n_samples_confidence_interval collections defaultdictlist_scores)rconfidence_interval n_samplesrrr__init__Ps zBootstrapAggregator.__init__cCs(t|D] \}}|j||qdS)zAdds a sample for future aggregation. Args: scores: Dict mapping score_type strings to a namedtuple object/class representing a score. N)six iteritemsr-append)rscores score_typerrrr add_scoreseszBootstrapAggregator.add_scorescspi}t|jD]-\}tt}||tfddtdD}t|d|d|dd||<q|S)zAggregates scores previously added using add_scores. Returns: A dict mapping score_type to AggregateScore objects. c3s*|]}dj|ddfVqdS)rN) __class__).0j percentilesr4rr s(z0BootstrapAggregator.aggregate..rr&r) r1r2r-npvstacktuple_bootstrap_resamplerr)rresultr5 score_matrix intervalsrr:r aggregateps  zBootstrapAggregator.aggregatecCst|j|jdf}t|jD](}tjjt|jd|jdd}||ddf}tj|dd||ddf<qd|j d}dt |dd|g}tj ||ddS) aPerforms bootstrap resampling on a matrix of scores. Args: matrix: A 2-d matrix of (sample, measure). Returns: A 2-d matrix of (bounds, measure). There are three bounds: low (row 0), mid (row 1) and high (row 2). Mid is always the mean, while low and high bounds are specified by self._confidence_interval (which defaults to 0.95 meaning it will return the 2.5th and 97.5th percentiles for a 95% confidence interval on the mean). r&r)sizeN)axisr>dg?) r?zerosr(shaperrandomchoicearangemeanr)array percentile)rmatrix sample_meani sample_idxsamplepercentile_deltaqrrrrBsz'BootstrapAggregator._bootstrap_resampleN)r$r%)r r r r r0r6rFrBrrrrr#<s   r#cCs$||dkrd||||SdS)z5Computes f-measure given precision and recall values.rr>gr)rrrrrrs r)r __future__rrrrr*typingrnumpyr?r1 six.movesr namedtuplerobjectABCMetarrr#rrrrrs&        e