def main():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i', '--input', metavar="INFILE", type=open,
dest="infile", help="source of analysis data")
a.add_argument('-v', '--verbose', action='store_true',
help="print verbosely while processing")
a.add_argument('-o', '--output', metavar="OUTFILE", dest="outfile",
help="print output into OUTFILE", type=FileType('w'))
a.add_argument('-x', '--statistics', metavar="STATFILE", dest="statfile",
help="print statistics to STATFILE", type=FileType('w'))
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True, accept=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
if not options.infile:
options.infile = stdin
if options.verbose:
print("analysing", options.infile.name)
if not options.outfile:
options.outfile = stdout
if options.verbose:
print("writing to", options.outfile.name)
if not options.statfile:
options.statfile = stdout
# statistics
realstart = perf_counter()
cpustart = process_time()
tokens = 0
unknowns = 0
for line in options.infile:
line = line
if not line or line == '':
continue
surfs = omorfi.tokenise(line)
for surf in surfs:
tokens += 1
anals = omorfi.analyse(surf)
print_analyses_vislcg3(surf, anals, options.outfile)
if len(anals) == 0 or (len(anals) == 1 and
'UNKNOWN' in anals[0][0]):
unknowns += 1
cpuend = process_time()
realend = perf_counter()
print("Tokens:", tokens, "Unknown:", unknowns, unknowns / tokens * 100,
"%", file=options.statfile)
print("CPU time:", cpuend - cpustart, "Real time:", realend - realstart,
file=options.statfile)
print("Tokens per timeunit:", tokens / (realend - realstart),
file=options.statfile)
exit(0)
def main():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i', '--input', metavar="INFILE", type=open,
dest="infile", help="source of analysis data")
a.add_argument('-v', '--verbose', action='store_true',
help="print verbosely while processing")
a.add_argument('-o', '--output', metavar="OUTFILE", dest="outfile",
help="print output into OUTFILE", type=FileType('w'))
a.add_argument('-x', '--statistics', metavar="STATFILE", dest="statfile",
help="print statistics to STATFILE", type=FileType('w'))
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True, accept=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
if not options.infile:
options.infile = stdin
if options.verbose:
print("analysing", options.infile.name)
if not options.outfile:
options.outfile = stdout
if options.verbose:
print("writing to", options.outfile.name)
if not options.statfile:
options.statfile = stdout
# statistics
realstart = perf_counter()
cpustart = process_time()
tokens = 0
unknowns = 0
for line in options.infile:
line = line
if not line or line == '':
continue
surfs = omorfi.tokenise(line)
for surf in surfs:
tokens += 1
anals = omorfi.analyse(surf)
print_analyses_vislcg3(surf, anals, options.outfile)
if len(anals) == 0 or (len(anals) == 1 and
'UNKNOWN' in anals[0][0]):
unknowns += 1
cpuend = process_time()
realend = perf_counter()
print("Tokens:", tokens, "Unknown:", unknowns, unknowns / tokens * 100,
"%", file=options.statfile)
print("CPU time:", cpuend - cpustart, "Real time:", realend - realstart,
file=options.statfile)
print("Tokens per timeunit:", tokens / (realend - realstart),
file=options.statfile)
exit(0)
def stream(text):
om = Omorfi()
om.load_from_dir('/usr/local/share/omorfi/', analyse=True)
for token in om.tokenise(text):
yield "%s\n" % token[0]
for analyse_res in om.analyse(token):
text, weight = analyse_res[:2]
if len(analyse_res) > 2:
rest = " ".join([str(x) for x in analyse_res[2:]])
else:
rest = ''
yield "%s %s %s\n" % (text, weight, rest)
yield "\n"
class FinnishParser:
def __init__(self):
self.omorfi = Omorfi()
self.omorfi.load_from_dir()
self.tokenizer = RegexpTokenizer(
'\w+\-\w+|\w+|\$[\d\.]+|\.\.\.|[",!\.\(\)]|\S+')
@staticmethod
def omorfi_to_base(omorfi_form):
return re.search(r"\[WORD_ID=(.*?)\]", omorfi_form).group(1)
@staticmethod
def omorfi_to_grammar(omorfi_form):
return re.sub(r"\[WORD_ID=.*?\]", "", omorfi_form)
def tokenize(self, text):
text = re.sub("\[\d+\]|\ufeff", "", text)
return self.tokenizer.tokenize(text)
def get_sentence_start_indexes(self, tokens):
start_indexes = []
sentence_ended = False
sentence_end_regex = r"\.\.\.|[\.!\?:;]"
for i, token in enumerate(tokens):
if re.match(sentence_end_regex, token):
sentence_ended = True
else:
if sentence_ended:
start_indexes.append(i)
sentence_ended = False
return start_indexes
def parse(self, text):
tokens = self.tokenize(text)
parsed_words = [self.analyse(t) for t in tokens]
sentence_start_indexes = self.get_sentence_start_indexes(tokens)
return parsed_words, tokens, sentence_start_indexes
def analyse(self, word):
omorfi_form = self.omorfi.analyse(word)
first_form = omorfi_form[0][0]
return AnalysedWord(self.omorfi_to_base(first_form),
self.omorfi_to_grammar(first_form))
def is_valid_word(self, word):
return word.grammar != "[GUESS=UNKNOWN][WEIGHT=inf]"
def main():
a = ArgumentParser()
a.add_argument('-f',
'--fsa',
metavar='FSAFILE',
required=True,
help="HFST's optimised lookup binary data for the "
"transducer to be applied")
options = a.parse_args()
omorfi = Omorfi()
omorfi.load_from_dir(options.fsa, analyse=True, accept=True)
tokens = omorfi.python_tokenise(WEIRD_TOK)
# Check tokens are in same order as text
start = 0
for token in tokens:
start = WEIRD_TOK.index(token['surf'], start)
def main():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-f',
'--fsa',
metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i',
'--input',
metavar="INFILE",
type=open,
dest="infile",
help="source of analysis data")
a.add_argument('-v',
'--verbose',
action='store_true',
help="print verbosely while processing")
a.add_argument('-o',
'--output',
metavar="OUTFILE",
dest="outfile",
help="print output into OUTFILE",
type=FileType('w'))
a.add_argument('-x',
'--statistics',
metavar="STATFILE",
dest="statfile",
help="print statistics to STATFILE",
type=FileType('w'))
a.add_argument('-O',
'--oracle',
action='store_true',
help="match to values in input when parsing if possible")
a.add_argument('-u',
'--udpipe',
metavar="UDPIPE",
help='use UDPIPE for additional guesses (experi-mental)')
a.add_argument('--hacks',
metavar='HACKS',
help="mangle anaelyses to match HACKS version of UD",
choices=['ftb'])
a.add_argument('--debug',
action='store_true',
help="print lots of debug info while processing")
options = a.parse_args()
if options.verbose:
print("Printing verbosely")
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True, guesser=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
if options.udpipe:
if options.verbose:
print("Loading udpipe", options.udpipe)
omorfi.load_udpipe(options.udpipe)
if not options.infile:
print("reading from <stdin>")
options.infile = stdin
if options.verbose:
print("analysing", options.infile.name)
if not options.outfile:
options.outfile = stdout
if options.verbose:
print("writing to", options.outfile.name)
if not options.statfile:
options.statfile = stdout
# statistics
realstart = perf_counter()
cpustart = process_time()
tokens = 0
unknowns = 0
sentences = 0
recognised_comments = [
'sent_id =', 'text =', 'doc-name:', 'sentence-text:'
]
for line in options.infile:
fields = line.strip().split('\t')
if len(fields) == 10:
# conllu is 10 field format
tokens += 1
try:
index = int(fields[0])
except ValueError:
if '-' in fields[0]:
# MWE
continue
elif '.' in fields[0]:
# a ghost
continue
else:
print("Cannot figure out token index",
fields[0],
file=stderr)
exit(1)
surf = fields[1]
anals = omorfi.analyse(surf)
if not anals or len(anals) == 0 or (len(anals) == 1
and 'UNKNOWN' in anals[0][0]):
unknowns += 1
anals = omorfi.guess(surf)
if anals and len(anals) > 0:
if options.debug:
debug_analyses_conllu(fields, index, surf, anals,
options.outfile, options.hacks)
elif options.oracle:
try_analyses_conllu(fields, index, surf, anals,
options.outfile, options.hacks)
else:
print_analyses_conllu(index, surf, anals[0],
options.outfile, options.hacks)
elif line.startswith('#'):
print(line.strip(), file=options.outfile)
recognised = False
for rec in recognised_comments:
if line.startswith('# ' + rec):
recognised = True
if not recognised and options.verbose:
print("Warning! Unrecognised comment line:", line, sep='\n')
elif not line or line.strip() == '':
# retain exactly 1 empty line between sents
print(file=options.outfile)
sentences += 1
else:
print("Error in conllu format:", line, sep='\n', file=stderr)
exit(1)
cpuend = process_time()
realend = perf_counter()
print("Tokens:", tokens, "Sentences:", sentences, file=options.statfile)
print("Unknowns / OOV:",
unknowns,
"=",
unknowns / tokens * 100 if tokens != 0 else 0,
"%",
file=options.statfile)
print("CPU time:",
cpuend - cpustart,
"Real time:",
realend - realstart,
file=options.statfile)
print("Tokens per timeunit:",
tokens / (realend - realstart),
file=options.statfile)
exit(0)
def main():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i', '--input', metavar="INFILE", type=open,
dest="infile", help="source of analysis data")
a.add_argument('-v', '--verbose', action='store_true',
help="print verbosely while processing")
a.add_argument('-o', '--output', metavar="OUTFILE", dest="outfile",
help="print output into OUTFILE", type=FileType('w'))
a.add_argument('-x', '--statistics', metavar="STATFILE", dest="statfile",
help="print statistics to STATFILE", type=FileType('w'))
a.add_argument('-O', '--output-format', metavar="OUTFORMAT",
default="moses",
help="format output for OUTFORMAT", choices=['moses', 'conllu'])
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True, accept=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
if not options.infile:
options.infile = stdin
if options.verbose:
print("analysing", options.infile.name)
if not options.outfile:
options.outfile = stdout
if options.verbose:
print("writing to", options.outfile.name)
if not options.statfile:
options.statfile = stdout
# statistics
realstart = perf_counter()
cpustart = process_time()
tokens = 0
lines = 0
if options.output_format == 'conllu':
print("# doc-name:", options.infile.name, file=options.outfile)
for line in options.infile:
line = line
lines += 1
if options.verbose and lines % 10000 == 0:
print(lines, "...")
if not line or line.rstrip('\n') == '':
continue
surfs = omorfi.tokenise(line)
tokens += len(surfs)
if options.output_format == 'moses':
print(' '.join([surf[0] for surf in surfs]), file=options.outfile)
else:
print("# sentence-text:", line.rstrip("\n"), file=options.outfile)
i = 1
for surf in surfs:
print(i, surf[0], "_", "_", "_", "_", "_", "_", "_",
surf[1],
sep="\t", file=options.outfile)
i += 1
if options.output_format == 'conllu':
print(file=options.outfile)
cpuend = process_time()
realend = perf_counter()
print("Lines:", lines, "Tokens:", tokens, "Ratio:", tokens / lines,
"tokens/line", file=options.statfile)
print("CPU time:", cpuend - cpustart, "Real time:", realend - realstart,
file=options.statfile)
print("Tokens per timeunit:", tokens / (realend - realstart),
"Lines per timeunit:", lines / (realend - realstart),
file=options.statfile)
exit(0)
def main():
"""Segment text in some formats."""
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i', '--input', metavar="INFILE", type=open,
dest="infile", help="source of analysis data")
a.add_argument('-v', '--verbose', action='store_true',
help="print verbosely while processing")
a.add_argument('-o', '--output', metavar="OUTFILE",
help="print segments into OUTFILE")
a.add_argument('-O', '--output-format', metavar="OFORMAT",
help="format output suitable for OFORMAT",
choices=["labels-tsv", "moses-factors", "segments"])
a.add_argument('--no-split-words', action="store_false", default=True,
dest="split_words",
help="split on word boundaries")
a.add_argument('--no-split-new-words', action="store_false", default=True,
dest="split_new_words",
help="split on new word boundaries (prev. unattested compounds)")
a.add_argument('--no-split-morphs', action="store_false", default=True,
dest="split_morphs",
help="split on morph boundaries")
a.add_argument('--split-derivs', action="store_true", default=False,
help="split on derivation boundaries")
a.add_argument('--split-nonwords', action="store_true", default=False,
help="split on other boundaries")
a.add_argument('--segment-marker', default='→ ←', metavar='SEG',
help="mark segment boundaries with SEG")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("Reading automata dir", options.fsa)
omorfi.load_from_dir(options.fsa, segment=True,
labelsegment=True)
else:
if options.verbose:
print("Searching for automata everywhere...")
omorfi.load_from_dir(labelsegment=True, segment=True)
if options.infile:
infile = options.infile
else:
options.infile = stdin
infile = stdin
if options.output:
outfile = open(options.output, 'w')
else:
options.output = "<stdout>"
outfile = stdout
if options.segment_marker is None:
if options.verbose:
print("Default segment marker is → ←")
options.segment_marker = '→ ←'
if options.verbose:
print("reading from", options.infile.name)
if options.verbose:
print("writign to", options.output)
linen = 0
for line in infile:
line = line.strip()
linen += 1
if options.verbose and linen % 10000 == 0:
print(linen, '...')
if not line or line == '':
print(file=outfile)
continue
tokens = omorfi.tokenise(line)
for token in tokens:
segments = omorfi.segment(token[0])
labelsegments = omorfi.labelsegment(token[0])
if options.output_format == 'moses-factors':
print_moses_factor_segments(
segments, labelsegments, token[0], outfile, options)
elif options.output_format == 'segments':
print_segments(segments, labelsegments, token[0], outfile,
options)
print(file=outfile)
exit(0)
def main():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i', '--input', metavar="INFILE", type=open,
dest="infile", help="source of analysis data")
a.add_argument('-v', '--verbose', action='store_true',
help="print verbosely while processing")
a.add_argument('-o', '--output', metavar="OUTFILE", dest="outfile",
help="print output into OUTFILE", type=FileType('w'))
a.add_argument('-x', '--statistics', metavar="STATFILE", dest="statfile",
help="print statistics to STATFILE", type=FileType('w'))
a.add_argument('-O', '--oracle', action='store_true',
help="match to values in input when parsing if possible")
a.add_argument('--hacks', metavar='HACKS',
help="mangle anaelyses to match HACKS version of UD",
choices=['ftb'])
a.add_argument('--debug', action='store_true',
help="print lots of debug info while processing")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
if not options.infile:
print("reading from <stdin>")
options.infile = stdin
if options.verbose:
print("analysing", options.infile.name)
if not options.outfile:
options.outfile = stdout
if options.verbose:
print("writing to", options.outfile.name)
if not options.statfile:
options.statfile = stdout
# statistics
realstart = perf_counter()
cpustart = process_time()
tokens = 0
unknowns = 0
sentences = 0
for line in options.infile:
fields = line.strip().split('\t')
if len(fields) == 10:
# conllu is 10 field format
tokens += 1
try:
index = int(fields[0])
except ValueError:
if '-' in fields[0]:
continue
else:
print(
"Cannot figure out token index", fields[0], file=stderr)
exit(1)
surf = fields[1]
anals = omorfi.analyse(surf)
if anals and len(anals) > 0:
if options.debug:
debug_analyses_conllu(
fields, index, surf, anals, options.outfile, options.hacks)
elif options.oracle:
try_analyses_conllu(fields, index, surf, anals,
options.outfile, options.hacks)
else:
print_analyses_conllu(index, surf, anals[0],
options.outfile, options.hacks)
if not anals or len(anals) == 0 or (len(anals) == 1 and
'UNKNOWN' in anals[0][0]):
unknowns += 1
elif line.startswith('# doc-name:') or line.startswith('# sentence-text:'):
# these comments I know need to be retained as-is
print(line.strip(), file=options.outfile)
elif line.startswith('#'):
# unknown comment
print(line.strip(), file=options.outfile)
if options.verbose:
print("Warning! Unrecognised comment line:", line, sep='\n')
elif not line or line.strip() == '':
# retain exactly 1 empty line between sents
print(file=options.outfile)
sentences += 1
else:
print("Error in conllu format:", line, sep='\n', file=stderr)
exit(1)
cpuend = process_time()
realend = perf_counter()
print("Tokens:", tokens, "Sentences:", sentences,
file=options.statfile)
print("Unknowns / OOV:", unknowns, "=",
unknowns / tokens * 100 if tokens != 0 else 0,
"%", file=options.statfile)
print("CPU time:", cpuend - cpustart, "Real time:", realend - realstart,
file=options.statfile)
print("Tokens per timeunit:", tokens / (realend - realstart),
file=options.statfile)
exit(0)
def main():
a = ArgumentParser()
a.add_argument(
'-f',
'--fsa',
metavar='FSAFILE',
required=True,
help=
"HFST's optimised lookup binary data for the transducer to be applied")
a.add_argument('-i',
'--input',
metavar="INFILE",
type=open,
required=True,
dest="infile",
help="source of analysis data")
a.add_argument('-o',
'--output',
metavar="outFILE",
type=FileType('w'),
required=True,
dest="outfile",
help="log file name")
a.add_argument('-v',
'--verbose',
action="store_true",
default=False,
help="Print verbosely while processing")
a.add_argument('-c',
'--count',
metavar="FREQ",
default=0,
help="test only word-forms with frequency higher than FREQ")
a.add_argument('-t',
'--threshold',
metavar='THOLD',
default=99,
type=int,
help="require THOLD % coverage or exit 1 (for testing)")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True, accept=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
# statistics
tokens = 0
uniqs = 0
found_tokens = 0
found_uniqs = 0
missed_tokens = 0
missed_uniqs = 0
# for make check target
realstart = perf_counter()
cpustart = process_time()
for line in options.infile:
fields = line.strip().replace(' ', '\t', 1).split('\t')
if len(fields) < 2:
print("ERROR: Skipping line", fields, file=stderr)
continue
freq = int(fields[0])
if freq < int(options.count):
break
surf = fields[1]
tokens += freq
uniqs += 1
if options.verbose:
print(tokens, "(", freq, ')...', end='\r')
anals = omorfi.analyse(surf)
if len(anals) > 0 and "GUESS=UNKNOWN" not in anals[0][0]:
found_tokens += freq
found_uniqs += 1
else:
missed_tokens += freq
missed_uniqs += 1
print(freq, surf, "? (missed)", sep="\t", file=options.outfile)
if options.verbose:
print()
cpuend = process_time()
realend = perf_counter()
print("cpu time: ", cpuend - cpustart, "real time:", realend - realstart)
print("Tokens", "Matches", "Misses", "%", sep="\t")
print(tokens,
found_tokens,
missed_tokens,
found_tokens / tokens * 100 if tokens != 0 else 0,
sep="\t")
print("Uniqs", "Matches", "Misses", "%", sep="\t")
print(uniqs,
found_uniqs,
missed_uniqs,
found_uniqs / uniqs * 100 if uniqs != 0 else 0,
sep="\t")
if tokens == 0 or (found_tokens / tokens * 100 < options.threshold):
print("needs to have",
options.threshold,
"% non-unique matches to pass regress test\n",
file=stderr)
exit(1)
else:
exit(0)
def main():
"""Segment text in some formats."""
a = ArgumentParser()
a.add_argument('-f',
'--fsa',
metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i',
'--input',
metavar="INFILE",
type=open,
dest="infile",
help="source of analysis data")
a.add_argument('-v',
'--verbose',
action='store_true',
help="print verbosely while processing")
a.add_argument('-o',
'--output',
metavar="OUTFILE",
help="print segments into OUTFILE")
a.add_argument('-O',
'--output-format',
metavar="OFORMAT",
help="format output suitable for OFORMAT",
choices=["labels-tsv", "moses-factors", "segments"])
a.add_argument('--split-words',
action="store_true",
default=True,
help="split on word boundaries")
a.add_argument(
'--split-new-words',
action="store_true",
default=True,
help="split on new word boundaries (prev. unattested compounds)")
a.add_argument('--split-morphs',
action="store_true",
default=True,
help="split on morph boundaries")
a.add_argument('--split-derivs',
action="store_true",
default=False,
help="split on derivation boundaries")
a.add_argument('--split-nonwords',
action="store_true",
default=True,
help="split on other boundaries")
a.add_argument('--segment-marker',
default=' ',
metavar='SEG',
help="mark segment boundaries with SEG")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("Reading automata dir", options.fsa)
omorfi.load_from_dir(options.fsa, segment=True, labelsegment=True)
else:
if options.verbose:
print("Searching for automata everywhere...")
omorfi.load_from_dir(labelsegment=True, segment=True)
if options.infile:
infile = options.infile
else:
infile = stdin
if options.output:
outfile = open(options.output, 'w')
else:
outfile = stdout
if options.verbose:
print("reading from", options.infile.name)
if options.verbose:
print("writign to", options.output)
linen = 0
for line in infile:
line = line.strip()
linen += 1
if options.verbose and linen % 10000 == 0:
print(linen, '...')
if not line or line == '':
continue
surfs = omorfi.tokenise(line)
for surf in surfs:
segments = omorfi.segment(surf)
labelsegments = omorfi.labelsegment(surf)
if options.output_format == 'moses-factors':
print_moses_factor_segments(segments, labelsegments, surf,
outfile)
elif options.output_format == 'segments':
print_segments(segments, labelsegments, surf, outfile, options)
print(file=outfile)
exit(0)
def main():
"""Preprocess text for moses factored modeling."""
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i', '--input', metavar="INFILE", type=open,
dest="infile", help="source of analysis data")
a.add_argument('-v', '--verbose', action='store_true',
help="print verbosely while processing")
a.add_argument('-o', '--output', metavar="OUTFILE",
help="print factors into OUTFILE")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("Reading automata dir", options.fsa)
omorfi.load_from_dir(options.fsa)
else:
if options.verbose:
print("Searching for automata everywhere...")
omorfi.load_from_dir()
if options.infile:
infile = options.infile
else:
infile = stdin
if options.output:
outfile = open(options.output, 'w')
else:
outfile = stdout
if options.verbose:
print("reading from", options.infile.name)
if options.verbose:
print("writign to", options.output)
re_lemma = re.compile("\[WORD_ID=([^]]*)\]")
re_pos = re.compile("\[POS=([^]]*)\]")
re_mrd = re.compile("\[([^=]*)=([^]]*)]")
linen = 0
for line in infile:
line = line.strip()
linen += 1
if options.verbose and linen % 10000 == 0:
print(linen, '...')
if not line or line == '':
continue
surfs = line.split()
for surf in surfs:
anals = omorfi.analyse(surf)
segments = omorfi.segment(surf)
pos_matches = re_pos.finditer(anals[0][0])
pos = "UNK"
mrds = []
lemmas = []
for pm in pos_matches:
pos = pm.group(1)
lemma_matches = re_lemma.finditer(anals[0][0])
for lm in lemma_matches:
lemmas += [lm.group(1)]
mrd_matches = re_mrd.finditer(anals[0][0])
for mm in mrd_matches:
if mm.group(1) == 'WORD_ID':
mrds = []
elif mm.group(1) == 'WEIGHT':
pass
else:
mrds += [mm.group(2)]
stemfixes = segments[0][0][
segments[0][0].rfind("{STUB}"):].replace("{STUB}", "")
if '{' in stemfixes:
morphs = stemfixes[stemfixes.find("{"):].replace("{MB}", ".")
else:
morphs = '0'
print(surf, '+'.join(lemmas), pos, '.'.join(mrds),
morphs, sep='|', end=' ', file=outfile)
print(file=outfile)
exit(0)
def stream(text):
om = Omorfi()
om.load_from_dir('/usr/local/share/omorfi/', lemmatise=True)
for token in om.tokenise(text):
yield " ".join(map(lambda x: str(x), om.lemmatise(token[0])))
def main():
a = ArgumentParser()
a.add_argument('-f', '--fsa', metavar='FSAFILE', required=True,
help="HFST's optimised lookup binary data for the transducer to be applied")
a.add_argument('-i', '--input', metavar="INFILE", type=open, required=True,
dest="infile", help="source of analysis data")
a.add_argument('-o', '--output', metavar="outFILE", type=FileType('w'),
required=True,
dest="outfile", help="log file name")
a.add_argument('-v', '--verbose', action="store_true", default=False,
help="Print verbosely while processing")
a.add_argument('-c', '--count', metavar="FREQ", default=0,
help="test only word-forms with frequency higher than FREQ")
a.add_argument('-t', '--threshold', metavar='THOLD', default=99, type=int,
help="require THOLD % coverage or exit 1 (for testing)")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("reading language models in", options.fsa)
omorfi.load_from_dir(options.fsa, analyse=True, accept=True)
else:
if options.verbose:
print("reading language models in default dirs")
omorfi.load_from_dir()
# statistics
tokens = 0
uniqs = 0
found_tokens = 0
found_uniqs = 0
missed_tokens = 0
missed_uniqs = 0
# for make check target
realstart = perf_counter()
cpustart = process_time()
for line in options.infile:
fields = line.strip().replace(' ', '\t', 1).split('\t')
if len(fields) < 2:
print("ERROR: Skipping line", fields, file=stderr)
continue
freq = int(fields[0])
if freq < int(options.count):
break
surf = fields[1]
tokens += freq
uniqs += 1
if options.verbose:
print(tokens, "(", freq, ')...', end='\r')
anals = omorfi.analyse(surf)
if len(anals) > 0 and "GUESS=UNKNOWN" not in anals[0][0]:
found_tokens += freq
found_uniqs += 1
else:
missed_tokens += freq
missed_uniqs += 1
print(freq, surf, "? (missed)", sep="\t", file=options.outfile)
if options.verbose:
print()
cpuend = process_time()
realend = perf_counter()
print("cpu time: ", cpuend - cpustart,
"real time:", realend - realstart)
print("Tokens", "Matches", "Misses", "%", sep="\t")
print(tokens, found_tokens, missed_tokens,
found_tokens / tokens * 100 if tokens != 0 else 0,
sep="\t")
print("Uniqs", "Matches", "Misses", "%", sep="\t")
print(uniqs, found_uniqs, missed_uniqs,
found_uniqs / uniqs * 100 if uniqs != 0 else 0,
sep="\t")
if tokens == 0 or (found_tokens / tokens * 100 < options.threshold):
print("needs to have", options.threshold,
"% non-unique matches to pass regress test\n",
file=stderr)
exit(1)
else:
exit(0)
def main():
"""Segment text in some formats."""
a = ArgumentParser()
a.add_argument('-f',
'--fsa',
metavar='FSAPATH',
help="Path to directory of HFST format automata")
a.add_argument('-i',
'--input',
metavar="INFILE",
type=open,
dest="infile",
help="source of analysis data")
a.add_argument('-v',
'--verbose',
action='store_true',
help="print verbosely while processing")
a.add_argument('-o',
'--output',
metavar="OUTFILE",
help="print segments into OUTFILE")
a.add_argument('-O',
'--output-format',
metavar="OFORMAT",
help="format output suitable for OFORMAT",
required=True,
choices=["moses-factors", "segments"])
a.add_argument('--no-split-words',
action="store_false",
default=True,
dest="split_words",
help="split on word boundaries")
a.add_argument(
'--no-split-new-words',
action="store_false",
default=True,
dest="split_new_words",
help="split on new word boundaries (prev. unattested compounds)")
a.add_argument('--no-split-morphs',
action="store_false",
default=True,
dest="split_morphs",
help="split on morph boundaries")
a.add_argument('--split-derivs',
action="store_true",
default=False,
help="split on derivation boundaries")
a.add_argument('--split-nonwords',
action="store_true",
default=False,
help="split on other boundaries")
a.add_argument('--segment-marker',
default='→ ←',
metavar='SEG',
help="mark segment boundaries with SEG")
a.add_argument('--show-ambiguous',
default=False,
metavar='ASEP',
help="separate ambiguous segmentations with SEG")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.fsa:
if options.verbose:
print("Reading automata dir", options.fsa)
omorfi.load_from_dir(options.fsa,
segment=True,
labelsegment=True,
accept=True)
else:
if options.verbose:
print("Searching for automata everywhere...")
omorfi.load_from_dir(labelsegment=True, segment=True, accept=True)
if not omorfi.can_segment:
print("Could not load segmenter(s), re-compile them or use -f option")
print()
print("To compile segmenter, use --enable-segmenter, and/or",
"--enable-labeled-segments")
exit(1)
if options.infile:
infile = options.infile
else:
options.infile = stdin
infile = stdin
if options.output:
outfile = open(options.output, 'w')
else:
options.output = "<stdout>"
outfile = stdout
if options.segment_marker is None:
if options.verbose:
print("Default segment marker is → ←")
options.segment_marker = '→ ←'
if options.verbose:
print("reading from", options.infile.name)
if options.verbose:
print("writign to", options.output)
linen = 0
for line in infile:
line = line.strip()
linen += 1
if options.verbose and linen % 10000 == 0:
print(linen, '...')
if not line or line == '':
print(file=outfile)
continue
tokens = omorfi.tokenise(line)
for token in tokens:
segments = omorfi.segment(token[0])
labelsegments = omorfi.labelsegment(token[0])
if options.output_format == 'moses-factors':
print_moses_factor_segments(segments, labelsegments, token[0],
outfile, options)
elif options.output_format == 'segments':
print_segments(segments, labelsegments, token[0], outfile,
options)
print(file=outfile)
exit(0)
