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)
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'))
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 get_omorfi():
"""
Gets an Omorfi instance with everything possible enabled. Reuses the
existing instance if already called once.
"""
from omorfi.omorfi import Omorfi
global _omorfi
if _omorfi is None:
_omorfi = Omorfi()
for var, fn in FSTS:
getattr(_omorfi, "load_" + var)(
"/usr/local/share/omorfi/omorfi." + fn
)
return _omorfi
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"
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)
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License version 3 as
# published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
from omorfi.omorfi import Omorfi
omorfi = Omorfi()
omorfi.load_analyser("/usr/local/share/omorfi/omorfi.analyse.hfst")
omorfi.load_generator("/usr/local/share/omorfi/omorfi.generate.hfst")
import settings
PROPERTIES = {
"nominatiivi": [("CASE", "NOM")],
"genetiivi": [("CASE", "GEN")],
"partitiivi": [("CASE", "PAR")],
"translatiivi": [("CASE", "TRA")],
"essiivi": [("CASE", "ESS")],
"inessiivi": [("CASE", "INE")],
"elatiivi": [("CASE", "ELA")],
"illatiivi": [("CASE", "ILL")],
"adessiivi": [("CASE", "ADE")],
def main():
global sent
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=str,
required=True,
dest="infile",
help="source of analysis data")
a.add_argument('-m',
'--master',
metavar="TSVFILE",
type=str,
required=True,
dest="tsvfile",
help="source of existing lexical data")
opts = a.parse_args()
if opts.infile:
test_corpora_files = [opts.infile]
else:
test_corpora_files = glob("*.text")
# hard-coded logs for now
lemma_log = open('missing_word_ids.log', 'w')
case_log = open('missing_nominal_cases.log', 'w')
comp_log = open('missing_comparatives.log', 'w')
adposition_log = open('adposition_complements.log', 'w')
adposition_stats = open('adposition_complements_full.log', 'w')
adjective_log = open('adjective_agreements.log', 'w')
omorfi = Omorfi()
omorfi.load_filename(opts.fsa)
gather_lemmas(open(opts.tsvfile))
test_corpora = list()
for test_corpus_file in test_corpora_files:
try:
test_corpora.append(open(test_corpus_file))
except IOError as ioe:
print("Failed to open corpus ", test_corpus_file, ":", ioe)
for test_corpus in test_corpora:
print('lines from', test_corpus.name)
linen = 0
for line in test_corpus:
linen += 1
if (linen % 200000) == 0:
print(
linen,
"...! Time to reload everything because memory is leaking very badly indeed!"
)
sent = list()
omorfi = None
omorfi = Omorfi()
omorfi.load_filename(opts.fsa)
gc.collect()
if (linen % 1000) == 0:
print(linen, "...", end='\r')
for punct in "\".,:;?!()":
line = line.replace(punct, " " + punct + " ")
for token in line.split():
analyses = omorfi.analyse(token)
add_to_sent(analyses, token)
stat_word_ids(token, analyses)
stat_nominal_cases(token, analyses, case_log)
stat_adjective_comps(token, analyses, comp_log)
print("Testing statistics")
test_zero_lemmas(lemma_log)
test_zero_cases(case_log)
test_zero_comps(comp_log)
# test_case_deviations()
test_adposition_complements(adposition_log)
test_adjective_agreements(adjective_log)
print("Writing accurate statistics")
print_adposition_stats(adposition_stats)
print_lemma_stats(open('lemmas.freqs', 'w'))
print_case_stats(open('cases.freqs', 'w'))
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():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-a', '--analyser', metavar='AFILE',
help="load tokeniser model from (analyser) AFILE",
required=True)
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', 'json', 'ftb3'])
options = a.parse_args()
omorfi = Omorfi(options.verbose)
if options.analyser:
if options.verbose:
print("reading language model", options.analyser)
omorfi.load_analyser(options.analyser)
else:
print("analyser is needed for tokenisation", file=stderr)
exit(1)
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("# new doc id=", 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['surf'] for surf in surfs]), file=options.outfile)
elif options.output_format == 'json':
print(json.encode(surfs))
elif options.output_format == 'conllu':
print("# sent_id =", lines, file=options.outfile)
print("# text =", line.rstrip("\n"), file=options.outfile)
i = 1
for surf in surfs:
print(i, surf['surf'], "_", "_", "_", "_", "_", "_", "_",
format_misc_ud(surf),
sep="\t", file=options.outfile)
i += 1
elif options.output_format == 'ftb3':
print("<s><loc file=\"", options.infile.name, "\" line=\"",
lines, "\" />", file=options.outfile, sep="")
i = 1
for surf in surfs:
print(i, surf['surf'], "_", "_", "_", "_", "_", "_", "_", "_",
sep="\t", file=options.outfile)
i += 1
print("</s>", file=options.outfile)
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():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-a',
'--analyser',
metavar='AFILE',
help="load analyser model from AFILE")
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('-F',
'--format',
metavar="INFORMAT",
default='text',
help="read input using INFORMAT tokenisation",
choices=['text', 'vislcg', 'conllu'])
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.analyser:
if options.verbose:
print("reading analyser model", options.analyser)
omorfi.load_analyser(options.analyser)
else:
print("analyser is required to vislcg", file=stderr)
exit(4)
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:
if options.outfile == stdout:
options.statfile = stdout
else:
options.statfile = stderr
# statistics
realstart = perf_counter()
cpustart = process_time()
tokens = 0
unknowns = 0
last = None
for line in options.infile:
surfs = []
if options.format == 'vislcg':
surfs = get_line_tokens_vislcg(line, last)
elif options.format == 'text':
surfs = get_line_tokens(line, omorfi)
elif options.format == 'conllu':
surfs = get_line_tokens_conllu(line, last)
else:
print("input format missing implementation",
options.format,
file=stderr)
exit(2)
for surf in surfs:
if 'conllu_form' in surf:
# skip conllu special forms in input for now:
# (ellipsis and MWE magics)
continue
elif 'surf' in surf:
tokens += 1
anals = omorfi.analyse(surf)
if len(anals) == 0 or (len(anals) == 1
and 'UNKNOWN' in anals[0]['anal']):
unknowns += 1
anals = omorfi.guess(surf)
print_analyses_vislcg3(surf, anals, options.outfile)
elif 'comment' in surf:
if surf['comment'].startswith(';') or \
surf['comment'].startswith('\t'):
continue
else:
print(surf['comment'], file=options.outfile)
elif 'error' in surf:
print(surf['error'], file=stderr)
exit(2)
last = surf
cpuend = process_time()
realend = perf_counter()
print("# Tokens:",
tokens,
"\n# Unknown:",
unknowns,
unknowns / tokens * 100,
"%",
file=options.statfile)
print("# CPU time:",
cpuend - cpustart,
"\n# Real time:",
realend - realstart,
file=options.statfile)
print("# Tokens per timeunit:",
tokens / (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('--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():
"""Invoke a simple CLI analyser."""
a = ArgumentParser()
a.add_argument('-a',
'--analyser',
metavar='AFILE',
help="read analyser model from AFILE",
required=True)
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('-X',
'--frequencies',
metavar="FREQDIR",
help="read frequencies from FREQDIR/*.freqs")
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.analyser:
if options.verbose:
print("reading analyser model", options.analyser)
omorfi.load_analyser(options.analyser)
else:
print("analyser is needed to ftb3", file=stderr)
exit(4)
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
lexprobs = None
tagprobs = None
if options.frequencies:
with open(options.frequencies + '/lexemes.freqs') as lexfile:
omorfi.load_lexical_frequencies(lexfile)
with open(options.frequencies + '/omors.freqs') as omorfile:
omorfi.load_omortag_frequencies(omorfile)
# 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:
# ftb is 10 field format
tokens += 1
try:
index = int(fields[0])
except ValueError:
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 'OOV' in anals[0]):
unknowns += 1
anals = omorfi.guess(surf)
if anals and len(anals) > 0:
if options.oracle:
try_analyses_ftb(fields, index, surf, anals,
options.outfile)
else:
print_analyses_ftb(index, surf, anals[0], options.outfile)
else:
print("Failed:", fields)
exit(1)
elif line.startswith('<') and line.rstrip().endswith('>'):
print(line.strip(), file=options.outfile)
elif not line or line.strip() == '':
# retain exactly 1 empty line between sents
print(file=options.outfile)
sentences += 1
else:
print("Error in ftb3 format: '", line, "'", 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 __init__(self):
self.omorfi = Omorfi()
self.omorfi.load_from_dir()
self.tokenizer = RegexpTokenizer(
'\w+\-\w+|\w+|\$[\d\.]+|\.\.\.|[",!\.\(\)]|\S+')
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():
"""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():
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():
global sent
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=str, required=True,
dest="infile", help="source of analysis data")
a.add_argument(
'-m', '--master', metavar="TSVFILE", type=str, required=True,
dest="tsvfile", help="source of existing lexical data")
opts = a.parse_args()
if opts.infile:
test_corpora_files = [opts.infile]
else:
test_corpora_files = glob("*.text")
# hard-coded logs for now
lemma_log = open('missing_word_ids.log', 'w')
case_log = open('missing_nominal_cases.log', 'w')
comp_log = open('missing_comparatives.log', 'w')
adposition_log = open('adposition_complements.log', 'w')
adposition_stats = open('adposition_complements_full.log', 'w')
adjective_log = open('adjective_agreements.log', 'w')
omorfi = Omorfi()
omorfi.load_filename(opts.fsa)
gather_lemmas(open(opts.tsvfile))
test_corpora = list()
for test_corpus_file in test_corpora_files:
try:
test_corpora.append(open(test_corpus_file))
except IOError as ioe:
print("Failed to open corpus ", test_corpus_file, ":", ioe)
for test_corpus in test_corpora:
print('lines from', test_corpus.name)
linen = 0
for line in test_corpus:
linen += 1
if (linen % 200000) == 0:
print(
linen, "...! Time to reload everything because memory is leaking very badly indeed!")
sent = list()
omorfi = None
omorfi = Omorfi()
omorfi.load_filename(opts.fsa)
gc.collect()
if (linen % 1000) == 0:
print(linen, "...", end='\r')
for punct in "\".,:;?!()":
line = line.replace(punct, " " + punct + " ")
for token in line.split():
analyses = omorfi.analyse(token)
add_to_sent(analyses, token)
stat_word_ids(token, analyses)
stat_nominal_cases(token, analyses, case_log)
stat_adjective_comps(token, analyses, comp_log)
print("Testing statistics")
test_zero_lemmas(lemma_log)
test_zero_cases(case_log)
test_zero_comps(comp_log)
# test_case_deviations()
test_adposition_complements(adposition_log)
test_adjective_agreements(adjective_log)
print("Writing accurate statistics")
print_adposition_stats(adposition_stats)
print_lemma_stats(open('lemmas.freqs', 'w'))
print_case_stats(open('cases.freqs', 'w'))
exit(0)
def main():
"""Preprocess text for moses factored modeling."""
a = ArgumentParser()
a.add_argument('-a', '--analyser', metavar='AFILE',
help="load analyser model from AFILE", required=True)
a.add_argument('-s', '--segmenter', metavar='SFILE',
help="load segmenter model from SFILE", required=True)
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.analyser:
if options.verbose:
print("Reading analyser model", options.analyser)
omorfi.load_analyser(options.analyser)
else:
print("at least analyser file is needed", file=stderr)
exit(1)
if options.segmenter:
if options.verbose:
print("Reading segmenter model", options.segmenter)
omorfi.load_segmenter(options.segmenter)
else:
print("at least segmenter file is needed", file=stderr)
exit(1)
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", infile.name)
if options.verbose:
print("writign to", outfile.name)
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
tokens = omorfi.tokenise_sentence(line)
for token in tokens:
if not token.surf:
continue
anals = omorfi.analyse(token)
pos = "X"
mrds = ["?"]
lemmas = [token.surf]
if anals:
anal = token.get_best()
pos = anal.get_upos()
mrds = anal.get_ufeats()
lemmas = anal.get_lemmas()
segments = omorfi.segment(token)
morphs = "0"
if segments:
segment = token.get_best_segments()
if segment:
parts = segment.get_segments()
morphs = ".".join(parts)
else:
morphs = token.surf
print(token.surf, '+'.join(lemmas), pos, '.'.join(mrds),
morphs, sep='|', end=' ', file=outfile)
print(file=outfile)
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():
a = ArgumentParser()
a.add_argument('-a',
'--analyser',
metavar='FSAFILE',
required=True,
help="load analyser from FSAFILE")
a.add_argument('-i',
'--input',
metavar="INFILE",
type=open,
dest="infile",
help="source of analysis data")
a.add_argument('-o',
'--output',
metavar="OUTFILE",
type=FileType('w'),
dest="outfile",
help="log outputs to OUTFILE")
a.add_argument('-X',
'--statistics',
metavar="STATFILE",
type=FileType('w'),
dest="statfile",
help="statistics")
a.add_argument('-v',
'--verbose',
action="store_true",
default=False,
help="Print verbosely while processing")
a.add_argument('-C',
'--no-casing',
action="store_true",
default=False,
help="Do not try to recase input and output when matching")
a.add_argument('-f',
'--format',
metavar="FORMAT",
help="use FORMAT formatter to compare analyses",
choices=["coverage", "ftb3.1"],
default="coverage")
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,
help="if coverage is less than THOLD exit with error")
options = a.parse_args()
omorfi = Omorfi(options.verbose)
try:
if options.analyser:
if options.verbose:
print("reading analyser from", options.analyser)
omorfi.load_analyser(options.analyser)
if not options.infile:
options.infile = stdin
print("reading from <stdin>")
if not options.statfile:
options.statfile = stdout
if not options.outfile:
options.outfile = stdout
except IOError:
print("Could not process file", options.analyser, file=stderr)
exit(2)
# basic statistics
covered = 0
full_matches = 0
lemma_matches = 0
anal_matches = 0
no_matches = 0
no_results = 0
lines = 0
# for make check target
threshold = options.threshold
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]
lemma = surf
analysis = surf
if options.format != 'coverage':
lemma = fields[2]
analysis = fields[3]
lines += freq
if options.verbose:
print(lines, '(', freq, ') ...', end='\r')
anals = omorfi.analyse(surf)
if not is_tokenlist_oov(anals):
covered += freq
else:
no_results += freq
print("OOV", surf, sep='\t', file=options.outfile)
found_anals = False
found_lemma = False
for anal in anals:
if options.format == 'ftb3.1':
anal_ftb3 = format_feats_ftb(anal)
lemma_ftb3 = '#'.join(get_lemmas(anal))
# hacks ftb3:
analysis = analysis.replace(" >>>", "")
if analysis == anal_ftb3:
found_anals = True
print("ANALHIT", analysis, anal_ftb3, file=options.outfile)
elif set(anal_ftb3.split()) == set(analysis.split()):
found_anals = True
print("PERMUTAHIT",
analysis,
anal_ftb3,
file=options.outfile)
else:
print("ANALMISS",
analysis,
anal_ftb3,
file=options.outfile)
if lemma == lemma_ftb3:
found_lemma = True
print("LEMMAHIT", lemma, lemma_ftb3, file=options.outfile)
elif lemma.replace('#', '') == lemma_ftb3.replace('#', ''):
found_lemma = True
print("LEMMARECOMP",
lemma,
lemma_ftb3,
file=options.outfile)
else:
print("LEMMAMISS", lemma, lemma_ftb3, file=options.outfile)
if options.format != 'coverage':
if not found_anals and not found_lemma:
no_matches += freq
print("NOHITS!", surf, sep='\t', file=options.outfile)
elif found_anals and found_lemma:
print("HIT", surf, sep='\t', file=options.outfile)
full_matches += freq
elif not found_anals:
anal_matches += freq
print("LEMMANOANAL", surf, sep='\t', file=options.outfile)
elif not found_lemma:
lemma_matches += freq
print("ANALNOLEMMA", surf, sep='\t', file=options.outfile)
else:
print("Logical error, kill everyone")
exit(13)
realend = perf_counter()
cpuend = process_time()
print("CPU time:", cpuend - cpustart, "real time:", realend - realstart)
print("Lines", "Covered", "OOV", sep="\t", file=options.statfile)
print(lines, covered, lines - covered, sep="\t", file=options.statfile)
print(lines / lines * 100 if lines != 0 else 0,
covered / lines * 100 if lines != 0 else 0,
(lines - covered) / lines * 100 if lines != 0 else 0,
sep="\t",
file=options.statfile)
if options.format == 'ftb3.1':
print("Lines",
"Matches",
"Lemma",
"Anals",
"Mismatch",
"No results",
sep="\t",
file=options.statfile)
print(lines,
full_matches,
lemma_matches,
anal_matches,
no_matches,
no_results,
sep="\t",
file=options.statfile)
print(lines / lines * 100 if lines != 0 else 0,
full_matches / lines * 100 if lines != 0 else 0,
lemma_matches / lines * 100 if lines != 0 else 0,
anal_matches / lines * 100 if lines != 0 else 0,
no_matches / lines * 100 if lines != 0 else 0,
no_results / lines * 100 if lines != 0 else 0,
sep="\t",
file=options.statfile)
if lines == 0:
print("Needs more than 0 lines to determine something", file=stderr)
exit(2)
elif options.format == 'ftb3.1' and \
(full_matches / lines * 100 <= int(options.threshold)):
print("needs to have",
threshold,
"% matches to pass regress test\n",
"please examine",
options.outfile.name,
"for regressions",
file=stderr)
exit(1)
elif options.format == 'coverage' and \
(covered / lines * 100 <= int(options.threshold)):
print("needs to have",
threshold,
"% coverage to pass regress test\n",
"please examine",
options.outfile.name,
"for regressions",
file=stderr)
exit(1)
else:
exit(0)
def main():
"""Segment text in some formats."""
a = ArgumentParser()
a.add_argument('-s',
'--segmenter',
metavar='SFILE',
help="load segmenter from SFILE",
required=True)
a.add_argument('-S',
'--labeller',
metavar='LSFILE',
help="load labelsegmenter from LSFILE",
required=True)
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.segmenter:
if options.verbose:
print("Reading segmenter", options.segmenter)
omorfi.load_segmenter(options.segmenter)
else:
print("segmenter is needed for segmenting", file=stderr)
exit(2)
if options.labeller:
if options.verbose:
print("Reading labelsegmenter", options.labeller)
omorfi.load_labelsegmenter(options.labeller)
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)
labelsegments = omorfi.labelsegment(token)
if options.output_format == 'moses-factors':
print_moses_factor_segments(segments, labelsegments, token,
outfile, options)
elif options.output_format == 'segments':
print_segments(segments, labelsegments, token, outfile,
options)
print(file=outfile)
exit(0)
def main():
"""Preprocess text for moses factored modeling."""
a = ArgumentParser()
a.add_argument('-a',
'--analyser',
metavar='AFILE',
help="load analyser model from AFILE",
required=True)
a.add_argument('-s',
'--segmenter',
metavar='SFILE',
help="load segmenter model from SFILE",
required=True)
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.analyser:
if options.verbose:
print("Reading analyser model", options.analyser)
omorfi.load_analyser(options.analyser)
else:
print("at least analyser file is needed", file=stderr)
exit(1)
if options.segmenter:
if options.verbose:
print("Reading segmenter model", options.segmenter)
omorfi.load_segmenter(options.segmenter)
else:
print("at least segmenter file is needed", file=stderr)
exit(1)
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", infile.name)
if options.verbose:
print("writign to", outfile.name)
re_lemma = re.compile("\[WORD_ID=([^]]*)\]")
re_pos = re.compile("\[UPOS=([^]]*)\]")
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]['anal'])
pos = "UNK"
mrds = []
lemmas = []
for pm in pos_matches:
pos = pm.group(1)
lemma_matches = re_lemma.finditer(anals[0]['anal'])
for lm in lemma_matches:
lemmas += [lm.group(1)]
mrd_matches = re_mrd.finditer(anals[0]['anal'])
for mm in mrd_matches:
if mm.group(1) == 'WORD_ID':
mrds = []
elif mm.group(1) == 'WEIGHT':
pass
else:
mrds += [mm.group(2)]
parts = segments[0]['segments']
if '{DB}' in parts:
suffixes = parts[parts.rfind('{DB}') + 4:]
elif '{WB}' in parts:
suffixes = parts[parts.rfind('{WB}') + 4:]
elif '{hyph?}' in parts:
suffixes = parts[parts.rfind('{hyph?}') + 6:]
else:
suffixes = "0"
morphs = suffixes[suffixes.find("{"):].replace("{MB}", ".")
print(surf,
'+'.join(lemmas),
pos,
'.'.join(mrds),
morphs,
sep='|',
end=' ',
file=outfile)
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('-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():
"""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)
