o pi@s,dZddlmZddZGdddeZdS)zFLibrary for computing the BLEU score Authors * Mirco Ravanelli 2021 ) MetricStatscCs&g}|D] }d|}||q|S)aMerge successive words into phrase, putting space between each word Arguments --------- sequences : list Each item contains a list, and this list contains a word sequence. Returns ------- The list contains phrase sequences.  )joinappend) sequencesresultsseqwordsr T/home/ubuntu/SoloSpeech/.venv/lib/python3.10/site-packages/speechbrain/utils/bleu.py merge_words s    r c@s6eZdZdZd ddZdddZdd d Zd d ZdS) BLEUStatsaA class for tracking BLEU (https://www.aclweb.org/anthology/P02-1040.pdf). Arguments --------- merge_words: bool Whether to merge the successive words to create sentences. max_ngram_order: int The maximum length of the ngrams to use for scoring. Example ------- >>> bleu = BLEUStats() >>> i2l = {0: 'a', 1: 'b'} >>> bleu.append( ... ids=['utterance1'], ... predict=[[0, 1, 1]], ... targets=[[[0, 1, 0]], [[0, 1, 1]], [[1, 1, 0]]], ... ind2lab=lambda batch: [[i2l[int(x)] for x in seq] for seq in batch], ... ) >>> stats = bleu.summarize() >>> stats['BLEU'] 0.0 TcCsVzddlm}Wn tytdYnw|||_||d|_g|_d|_dS)Nr)BLEUz^Please install sacrebleu (https://pypi.org/project/sacrebleu/) in order to use the BLEU metric)max_ngram_order) sacrebleu.metricsr ImportErrorprintclearr bleupredictstargets)selfr rrr r r __init__6s   zBLEUStats.__init__Ncs|j|dur|}fdd|D}|jr%t|}dd|D}|j||jdur5||_dSt|jt|ks@Jtt|jD] }|j|||qGdS)aUAdd stats to the relevant containers. * See MetricStats.append() Arguments --------- ids : list List of ids corresponding to utterances. predict : torch.tensor A predicted output, for comparison with the target output targets : list list of references (when measuring BLEU, one sentence could have more than one target translation). ind2lab : callable Callable that maps from indices to labels, operating on batches, for writing alignments. Ncsg|]}|qSr r .0tind2labr r Zz$BLEUStats.append..cSsg|]}t|qSr )r rr r r r^r )idsextendr rrlenrange)rr!predictrrir rr rFs    zBLEUStats.appendcCs|j|j|j}i}|j|d<|j|d<|j|j|d<|j|d<|j|d<|j|d<||_ ||_ |j d|j d<|durC|j |S|j S) zaSummarize the BLEU and return relevant statistics. * See MetricStats.summarize() rBPratiohyp_lenref_len precisions bleu_scoreN) r corpus_scorerrscorebpsys_lenr*r+scoressummary)rfieldr1detailsr r r summarizehs      zBLEUStats.summarizecCs |js|t|j|ddS)zoWrite all relevant info (e.g., error rate alignments) to file. * See MetricStats.write_stats() )fileN)r2r5rr1)r filestreamr r r write_statsszBLEUStats.write_stats)Tr)N)__name__ __module__ __qualname____doc__rrr5r8r r r r r s    " r N)r<speechbrain.utils.metric_statsrr r r r r r s