def train(
encoding: str,
model_path: str,
input_path: str or None,
tag_order: int,
emission_order: int,
suff_length: int,
rare_freq: int,
separator: str,
linesep: str,
): # todo verbose mode
"""Create a language model from an analysed corpora (and optionally from an existing model).
It performs on the given input which can be also the stdin.
:param encoding: The encoding of the corpora. If None, Python3 default will be used.
:param model_path: Path of the model file. If exists, it will be improved.
:param input_path: Path of the analysed corpora. If None, stdin will be used.
:param tag_order: # todo
:param emission_order: # todo
:param suff_length: # todo
:param rare_freq: # todo
:param separator: The sepatator character(s) inside the token. Default/traditionally: '#'.
:param linesep: The sepatator character(s) between the sentences. Default: newline.
"""
if input_path is not None:
source = open(input_path, encoding=encoding) # todo default encoding? (a Python3 okos)
else:
source = sys.stdin
trainer = Trainer(source, CorpusReader(StemmedTaggedTokenReader(separator, linesep)))
if os.path.isfile(model_path):
print("Reading model... ", file=sys.stderr)
ret_model = StandardSerializer.read_model(model_path)
print("Training model... ", file=sys.stderr)
ret_model = trainer.train_model(ret_model)
else:
print("Training model... ", file=sys.stderr)
ret_model = trainer.train(tag_order, emission_order, suff_length, rare_freq)
print(trainer.stat.stat(ret_model), file=sys.stderr)
print("Writing model... ", file=sys.stderr)
StandardSerializer.write_model(ret_model, model_path)
print("Done!", file=sys.stderr)
def create_tagger(
model_path: str,
analyser: str,
no_stemming: bool,
max_guessed: int,
beam_log_theta: float,
use_beam_search: bool,
conf: Configuration,
humor_path: str,
lex_path: str,
) -> POSTagger:
"""Create a tagger object with the given properties.
:param model_path:
:param analyser:
:param no_stemming:
:param max_guessed:
:param beam_log_theta:
:param use_beam_search:
:param conf:
:param humor_path:
:param lex_path:
:return: a tagger object.
"""
if analyser == PurePos.INTEGRATED_MA:
try:
ma = PurePos.load_humor(humor_path + "/bin/pyhumor/__init__.py", lex_path)
except FileNotFoundError:
print("Humor module not found. Not using any morphological analyzer.", file=sys.stderr)
ma = BaseMorphologicalAnalyser()
elif analyser == PurePos.NONE_MA:
ma = BaseMorphologicalAnalyser()
else:
print("Using morphological table at: {}.".format(analyser), file=sys.stderr)
ma = MorphologicalTable(open(analyser))
print("Reading model... ", file=sys.stderr)
rawmodel = StandardSerializer.read_model(model_path)
print("Compiling model... ", file=sys.stderr)
cmodel = rawmodel.compile(conf)
suff_log_theta = math.log(10)
if no_stemming:
tagger = POSTagger(cmodel, ma, beam_log_theta, suff_log_theta, max_guessed, use_beam_search)
else:
tagger = MorphTagger(cmodel, ma, beam_log_theta, suff_log_theta, max_guessed, use_beam_search)
return tagger
