wer.rar_compute wer_python_speech recognition_语言识别_语音识别

#!/usr/bin/env python2.7 # -*- coding: utf-8 -*- import argparse import logging import sys logging.basicConfig( format='%(levelname)s(%(filename)s:%(lineno)d): %(message)s') def levenshtein(u, v): prev = None curr = [0] + range(1, len(v) + 1) # Operations: (SUB, DEL, INS) prev_ops = None curr_ops = [(0, 0, i) for i in range(len(v) + 1)] for x in xrange(1, len(u) + 1): prev, curr = curr, [x] + ([None] * len(v)) prev_ops, curr_ops = curr_ops, [(0, x, 0)] + ([None] * len(v)) for y in xrange(1, len(v) + 1): delcost = prev[y] + 1 addcost = curr[y - 1] + 1 subcost = prev[y - 1] + int(u[x - 1] != v[y - 1]) curr[y] = min(subcost, delcost, addcost) if curr[y] == subcost: (n_s, n_d, n_i) = prev_ops[y - 1] curr_ops[y] = (n_s + int(u[x - 1] != v[y - 1]), n_d, n_i) elif curr[y] == delcost: (n_s, n_d, n_i) = prev_ops[y] curr_ops[y] = (n_s, n_d + 1, n_i) else: (n_s, n_d, n_i) = curr_ops[y - 1] curr_ops[y] = (n_s, n_d, n_i + 1) return curr[len(v)], curr_ops[len(v)] def load_file(fname, encoding): try: f = open(fname, 'r') data = [] for line in f: data.append(line.rstrip('\n').rstrip('\r').decode(encoding)) f.close() except: logging.error('Error reading file "%s"', fname) exit(1) return data def evaluate(reference, transcription): ref, hyp = [], [] ref.append(reference) hyp.append(transcription) wer_s, wer_i, wer_d, wer_n = 0, 0, 0, 0 cer_s, cer_i, cer_d, cer_n = 0, 0, 0, 0 sen_err = 0 for n in range(len(ref)): # update CER statistics _, (s, i, d) = levenshtein(ref[n], hyp[n]) cer_s += s cer_i += i cer_d += d cer_n += len(ref[n]) # update WER statistics _, (s, i, d) = levenshtein(ref[n].split(), hyp[n].split()) wer_s += s wer_i += i wer_d += d wer_n += len(ref[n].split()) # update SER statistics if s + i + d > 0: sen_err += 1 if cer_n > 0: print( 'CER: %g%%, WER: %g%%, SER: %g%%' % ( (100.0 * (cer_s + cer_i + cer_d)) / cer_n, (100.0 * (wer_s + wer_i + wer_d)) / wer_n, (100.0 * sen_err) / len(ref)) ) if __name__ == '__main__': evaluate('womenshihaopengyou', 'womenshijkjn') # parser = argparse.ArgumentParser( # description='Compute useful evaluation metrics (CER, WER, SER, ...)') # parser.add_argument( # '-r', '--reference', type=str, metavar='REF', default=None, # help='reference sentence or file') # parser.add_argument( # '-t', '--transcription', type=str, metavar='HYP', default=None, # help='transcription sentence or file') # parser.add_argument( # '-i', '--input_source', type=str, choices=('-', 'str', 'file'), # default='-', help=""""-" reads parallel sentences from the standard # input, "str" interprets `-r' and `-t' as sentences, and "file" # interprets `-r' and `-t' as two parallel files, with one sentence per # line (default: -)""") # parser.add_argument( # '-s', '--separator', type=str, metavar='SEP', default='\t', # help="""use this string to separate the reference and transcription # when reading from the standard input (default: \\t)""") # parser.add_argument( # '-e', '--encoding', type=str, metavar='ENC', default='utf-8', # help="""character encoding of the reference and transcription text # (default: utf-8)""") # args = parser.parse_args() # # if args.input_source != '-' and \ # (args.reference is None or args.transcription is None): # logging.error('Expected reference and transcription sources') # exit(1) # ref, hyp = [], [] # if args.input_source == 'str': # ref.append(args.reference.decode(args.encoding)) # hyp.append(args.transcription.decode(args.encoding)) # elif args.input_source == '-': # line_n = 0 # for line in sys.stdin: # line_n += 1 # line = line.rstrip('\n').rstrip('\r').decode(args.encoding) # fields = line.split(args.separator) # if len(fields) != 2: # logging.warning( # 'Line %d has %d fields but 2 were expected', # line_n, len(fields)) # continue # ref.append(fields[0]) # hyp.append(fields[1]) # elif args.input_source == 'file': # ref = load_file(args.reference, args.encoding) # hyp = load_file(args.transcription, args.encoding) # if len(ref) != len(hyp): # logging.error( # 'The number of reference and transcription sentences does not ' # 'match (%d vs. %d)', len(ref), len(hyp)) # exit(1) # else: # logging.error('INPUT FROM "%s" NOT IMPLEMENTED', args.input_source) # exit(1) # # wer_s, wer_i, wer_d, wer_n = 0, 0, 0, 0 # cer_s, cer_i, cer_d, cer_n = 0, 0, 0, 0 # sen_err = 0 # for n in range(len(ref)): # # update CER statistics # _, (s, i, d) = levenshtein(ref[n], hyp[n]) # cer_s += s # cer_i += i # cer_d += d # cer_n += len(ref[n]) # # update WER statistics # _, (s, i, d) = levenshtein(ref[n].split(), hyp[n].split()) # wer_s += s # wer_i += i # wer_d += d # wer_n += len(ref[n].split()) # # update SER statistics # if s + i + d > 0: # sen_err += 1 # # if cer_n > 0: # print( # 'CER: %g%%, WER: %g%%, SER: %g%%' % ( # (100.0 * (cer_s + cer_i + cer_d)) / cer_n, # (100.0 * (wer_s + wer_i + wer_d)) / wer_n, # (100.0 * sen_err) / len(ref)) # )