def main():
global online
logging.basicConfig(level=logging.INFO)
parser = argparse.ArgumentParser(description='Extract grammars from a compiled corpus.')
parser.add_argument('-c', '--config', required=True,
help='extractor configuration')
parser.add_argument('-g', '--grammars',
help='grammar output path')
parser.add_argument('-j', '--jobs', type=int, default=1,
help='number of parallel extractors')
parser.add_argument('-s', '--chunksize', type=int, default=10,
help='number of sentences / chunk')
parser.add_argument('-f', '--features', nargs='*', default=[],
help='additional feature definitions')
parser.add_argument('-o', '--online', action='store_true',
help='online grammar extraction')
parser.add_argument('-z', '--compress', action='store_true',
help='compress grammars with gzip')
parser.add_argument('-t', '--stream', action='store_true',
help='stream mode (see README.md)')
args = parser.parse_args()
if not (args.grammars or args.stream):
sys.stderr.write('Error: either -g/--grammars or -t/--stream required\n')
sys.exit(1)
if args.grammars and not os.path.exists(args.grammars):
os.mkdir(args.grammars)
for featdef in args.features:
if not featdef.endswith('.py'):
sys.stderr.write('Error: feature definition file <{}>'
' should be a python module\n'.format(featdef))
sys.exit(1)
online = args.online
stream = args.stream
start_time = monitor_cpu()
if args.jobs > 1:
if stream:
sys.stderr.write('Error: stream mode incompatible with multiple jobs\n')
sys.exit(1)
logging.info('Starting %d workers; chunk size: %d', args.jobs, args.chunksize)
pool = mp.Pool(args.jobs, make_extractor, (args,))
try:
for output in pool.imap(extract, enumerate(sys.stdin), args.chunksize):
print(output)
except KeyboardInterrupt:
pool.terminate()
else:
make_extractor(args)
if stream:
stream_extract()
else:
for output in map(extract, enumerate(sys.stdin)):
print(output)
stop_time = monitor_cpu()
logging.info("Overall extraction step took %f seconds", stop_time - start_time)
def main():
global online
logging.basicConfig(level=logging.INFO)
parser = argparse.ArgumentParser(
description='Extract grammars from a compiled corpus.')
parser.add_argument('-c',
'--config',
required=True,
help='extractor configuration')
parser.add_argument('-g',
'--grammars',
required=True,
help='grammar output path')
parser.add_argument('-j',
'--jobs',
type=int,
default=1,
help='number of parallel extractors')
parser.add_argument('-s',
'--chunksize',
type=int,
default=10,
help='number of sentences / chunk')
parser.add_argument('-f',
'--features',
nargs='*',
default=[],
help='additional feature definitions')
parser.add_argument('-o',
'--online',
action='store_true',
help='online grammar extraction')
parser.add_argument('-z',
'--compress',
action='store_true',
help='compress grammars with gzip')
args = parser.parse_args()
if not os.path.exists(args.grammars):
os.mkdir(args.grammars)
for featdef in args.features:
if not featdef.endswith('.py'):
sys.stderr.write('Error: feature definition file <{}>'
' should be a python module\n'.format(featdef))
sys.exit(1)
online = args.online
start_time = monitor_cpu()
if args.jobs > 1:
logging.info('Starting %d workers; chunk size: %d', args.jobs,
args.chunksize)
pool = mp.Pool(args.jobs, make_extractor, (args, ))
try:
for output in pool.imap(extract, enumerate(sys.stdin),
args.chunksize):
print(output)
except KeyboardInterrupt:
pool.terminate()
else:
make_extractor(args)
for output in map(extract, enumerate(sys.stdin)):
print(output)
stop_time = monitor_cpu()
logging.info("Overall extraction step took %f seconds",
stop_time - start_time)
def main():
preprocess_start_time = monitor_cpu()
sys.setrecursionlimit(sys.getrecursionlimit() * 100)
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger('cdec.sa.compile')
parser = argparse.ArgumentParser(description='Compile a corpus into a suffix array.')
parser.add_argument('--maxnt', '-n', type=int, default=2,
help='Maximum number of non-terminal symbols')
parser.add_argument('--maxlen', '-l', type=int, default=5,
help='Maximum number of terminals')
parser.add_argument('--maxsize', '-s', type=int, default=15,
help='Maximum rule span')
parser.add_argument('--mingap', '-g', type=int, default=1,
help='Minimum gap size')
parser.add_argument('--rank1', '-r1', type=int, default=100,
help='Number of pre-computed frequent patterns')
parser.add_argument('--rank2', '-r2', type=int, default=10,
help='Number of pre-computed super-frequent patterns)')
parser.add_argument('--loose', action='store_true',
help='Enable loose phrase extraction (default: tight)')
parser.add_argument('-c', '--config', default='/dev/stdout',
help='Output configuration')
parser.add_argument('-f', '--source',
help='Source language corpus')
parser.add_argument('-e', '--target',
help='Target language corpus')
parser.add_argument('-b', '--bitext',
help='Parallel text (source ||| target)')
parser.add_argument('-a', '--alignment', required=True,
help='Bitext word alignment')
parser.add_argument('-o', '--output', required=True,
help='Output path')
parser.add_argument('--online', action='store_true',
help='Compile data for online grammar extraction')
args = parser.parse_args()
if not ((args.source and args.target) or args.bitext):
parser.error('a parallel corpus is required\n'
'\tuse -f (source) with -e (target) or -b (bitext)')
param_names = ('max_len', 'max_nt', 'max_size', 'min_gap',
'rank1', 'rank2', 'tight_phrases')
params = (args.maxlen, args.maxnt, args.maxsize, args.mingap,
args.rank1, args.rank2, not args.loose)
if not os.path.exists(args.output):
os.mkdir(args.output)
f_sa_bin = os.path.join(args.output, 'f.sa.bin')
e_bin = os.path.join(args.output, 'e.bin')
precomp_file = 'precomp.{0}.{1}.{2}.{3}.{4}.{5}.bin'.format(*params)
precomp_bin = os.path.join(args.output, precomp_file)
a_bin = os.path.join(args.output, 'a.bin')
lex_bin = os.path.join(args.output, 'lex.bin')
# online only
bilex_file = os.path.join(args.output, 'bilex.gz')
config = cdec.configobj.ConfigObj(args.config, unrepr=True)
config['f_sa_file'] = os.path.abspath(f_sa_bin)
config['e_file'] = os.path.abspath(e_bin)
config['a_file'] = os.path.abspath(a_bin)
config['lex_file'] = os.path.abspath(lex_bin)
config['precompute_file'] = os.path.abspath(precomp_bin)
if args.online:
config['bilex_file'] = os.path.abspath(bilex_file)
start_time = monitor_cpu()
logger.info('Compiling source suffix array')
if args.bitext:
f_sa = cdec.sa.SuffixArray(from_text=args.bitext, side='source')
else:
f_sa = cdec.sa.SuffixArray(from_text=args.source)
f_sa.write_binary(f_sa_bin)
stop_time = monitor_cpu()
logger.info('Compiling source suffix array took %f seconds', stop_time - start_time)
start_time = monitor_cpu()
logger.info('Compiling target data array')
if args.bitext:
e = cdec.sa.DataArray(from_text=args.bitext, side='target')
else:
e = cdec.sa.DataArray(from_text=args.target)
e.write_binary(e_bin)
stop_time = monitor_cpu()
logger.info('Compiling target data array took %f seconds', stop_time - start_time)
start_time = monitor_cpu()
logger.info('Precomputing frequent phrases')
precompute(f_sa, *params).write_binary(precomp_bin)
stop_time = monitor_cpu()
logger.info('Compiling precomputations took %f seconds', stop_time - start_time)
start_time = monitor_cpu()
logger.info('Compiling alignment')
a = cdec.sa.Alignment(from_text=args.alignment)
a.write_binary(a_bin)
stop_time = monitor_cpu()
logger.info('Compiling alignment took %f seconds', stop_time - start_time)
start_time = monitor_cpu()
logger.info('Compiling bilexical dictionary')
lex = cdec.sa.BiLex(from_data=True, alignment=a, earray=e, fsarray=f_sa)
lex.write_binary(lex_bin)
stop_time = monitor_cpu()
logger.info('Compiling bilexical dictionary took %f seconds', stop_time - start_time)
if args.online:
start_time = monitor_cpu()
logger.info('Compiling online bilexical dictionary')
if args.bitext:
bilex = cdec.sa.online.Bilex()
bilex.add_bitext(args.alignment, args.bitext)
else:
bilex = cdec.sa.online.Bilex()
bilex.add_bitext(args.alignment, args.source, args.target)
bilex.write(bilex_file)
stop_time = monitor_cpu()
logger.info('Compiling online bilexical dictionary took %f seconds', stop_time - start_time)
# Write configuration
for name, value in zip(param_names, params):
config[name] = value
config.write()
preprocess_stop_time = monitor_cpu()
logger.info('Overall preprocessing step took %f seconds', preprocess_stop_time - preprocess_start_time)
def main():
preprocess_start_time = monitor_cpu()
sys.setrecursionlimit(sys.getrecursionlimit() * 100)
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger('cdec.sa.compile')
parser = argparse.ArgumentParser(
description='Compile a corpus into a suffix array.')
parser.add_argument('--maxnt',
'-n',
type=int,
default=2,
help='Maximum number of non-terminal symbols')
parser.add_argument('--maxlen',
'-l',
type=int,
default=5,
help='Maximum number of terminals')
parser.add_argument('--maxsize',
'-s',
type=int,
default=15,
help='Maximum rule span')
parser.add_argument('--mingap',
'-g',
type=int,
default=1,
help='Minimum gap size')
parser.add_argument('--rank1',
'-r1',
type=int,
default=100,
help='Number of pre-computed frequent patterns')
parser.add_argument('--rank2',
'-r2',
type=int,
default=10,
help='Number of pre-computed super-frequent patterns)')
parser.add_argument('--loose',
action='store_true',
help='Enable loose phrase extraction (default: tight)')
parser.add_argument('-c',
'--config',
default='/dev/stdout',
help='Output configuration')
parser.add_argument('-f', '--source', help='Source language corpus')
parser.add_argument('-e', '--target', help='Target language corpus')
parser.add_argument('-b',
'--bitext',
help='Parallel text (source ||| target)')
parser.add_argument('-a',
'--alignment',
required=True,
help='Bitext word alignment')
parser.add_argument('-o', '--output', required=True, help='Output path')
parser.add_argument('--online',
action='store_true',
help='Compile data for online grammar extraction')
args = parser.parse_args()
if not ((args.source and args.target) or args.bitext):
parser.error('a parallel corpus is required\n'
'\tuse -f (source) with -e (target) or -b (bitext)')
param_names = ('max_len', 'max_nt', 'max_size', 'min_gap', 'rank1',
'rank2', 'tight_phrases')
params = (args.maxlen, args.maxnt, args.maxsize, args.mingap, args.rank1,
args.rank2, not args.loose)
if not os.path.exists(args.output):
os.mkdir(args.output)
f_sa_bin = os.path.join(args.output, 'f.sa.bin')
e_bin = os.path.join(args.output, 'e.bin')
precomp_file = 'precomp.{0}.{1}.{2}.{3}.{4}.{5}.bin'.format(*params)
precomp_bin = os.path.join(args.output, precomp_file)
a_bin = os.path.join(args.output, 'a.bin')
lex_bin = os.path.join(args.output, 'lex.bin')
# online only
bilex_file = os.path.join(args.output, 'bilex.gz')
config = cdec.configobj.ConfigObj(args.config, unrepr=True)
config['f_sa_file'] = os.path.abspath(f_sa_bin)
config['e_file'] = os.path.abspath(e_bin)
config['a_file'] = os.path.abspath(a_bin)
config['lex_file'] = os.path.abspath(lex_bin)
config['precompute_file'] = os.path.abspath(precomp_bin)
if args.online:
config['bilex_file'] = os.path.abspath(bilex_file)
start_time = monitor_cpu()
logger.info('Compiling source suffix array')
if args.bitext:
f_sa = cdec.sa.SuffixArray(from_text=args.bitext, side='source')
else:
f_sa = cdec.sa.SuffixArray(from_text=args.source)
f_sa.write_binary(f_sa_bin)
stop_time = monitor_cpu()
logger.info('Compiling source suffix array took %f seconds',
stop_time - start_time)
start_time = monitor_cpu()
logger.info('Compiling target data array')
if args.bitext:
e = cdec.sa.DataArray(from_text=args.bitext, side='target')
else:
e = cdec.sa.DataArray(from_text=args.target)
e.write_binary(e_bin)
stop_time = monitor_cpu()
logger.info('Compiling target data array took %f seconds',
stop_time - start_time)
start_time = monitor_cpu()
logger.info('Precomputing frequent phrases')
precompute(f_sa, *params).write_binary(precomp_bin)
stop_time = monitor_cpu()
logger.info('Compiling precomputations took %f seconds',
stop_time - start_time)
start_time = monitor_cpu()
logger.info('Compiling alignment')
a = cdec.sa.Alignment(from_text=args.alignment)
a.write_binary(a_bin)
stop_time = monitor_cpu()
logger.info('Compiling alignment took %f seonds', stop_time - start_time)
start_time = monitor_cpu()
logger.info('Compiling bilexical dictionary')
lex = cdec.sa.BiLex(from_data=True, alignment=a, earray=e, fsarray=f_sa)
lex.write_binary(lex_bin)
stop_time = monitor_cpu()
logger.info('Compiling bilexical dictionary took %f seconds',
stop_time - start_time)
if args.online:
start_time = monitor_cpu()
logger.info('Compiling online bilexical dictionary')
if args.bitext:
bilex = cdec.sa.online.Bilex()
bilex.add_bitext(args.alignment, args.bitext)
else:
bilex = cdec.sa.online.Bilex()
bilex.add_bitext(args.alignment, args.source, args.target)
bilex.write(bilex_file)
stop_time = monitor_cpu()
logger.info('Compiling online bilexical dictionary took %f seconds',
stop_time - start_time)
# Write configuration
for name, value in zip(param_names, params):
config[name] = value
config.write()
preprocess_stop_time = monitor_cpu()
logger.info('Overall preprocessing step took %f seconds',
preprocess_stop_time - preprocess_start_time)
