def classify_file(train_inst_fns,
test_inst_fns,
out_fns=None,
log_fn=None,
clas_dir=None,
descriptor=None,
timbl=None,
options="",
log=False):
"""
Classify instances using Timbl
@param train_inst_fns: a list of instance filenames for training
@param test_inst_fns: a list of instance filenames for testing
@keyword out_fns: list of classifier output files to be created
@keyword log_fn: classifier log file to be created;
ignored if keyword log is false
@keyword clas_dir: directory for creating classifier output files; ignored
if out_fns is given
@keyword descriptor: a Descriptor instance, required to infer the feature
metrics for Timbl, unless a TimblFile is supplied; ignored if timbl is
supplied
@keyword timbl: a tailored TimblFile instance; notice that it must at
least set the verbosity options +vo, +vdb, +vdi, and the -m option to
specify that the administrative features must be ignored.
@keyword options: list of additional Timbl options, excluding -f, -m, +vo,
+vdb, +vdi
@keyword log: log Timbl's standard output and error streams to file
@return: list of Timbl output filenames
"""
if clas_dir:
makedirs(clas_dir)
if not timbl:
assert descriptor
timbl = TimblFile(default_opts=timbl_options_string(descriptor))
else:
# ignore descriptor
assert isinstance(timbl, TimblFile)
assert "+vo" in timbl.default_opts
assert "+vdb" in timbl.default_opts
assert "+vdi" in timbl.default_opts
assert "-m" in timbl.default_opts
return timbl.train_test_multi(train_inst_fns,
test_inst_fns,
out_fns=out_fns,
log_fn=log_fn,
options=options,
log=log,
out_dir=clas_dir)
def classify_file_cv(inst_fns,
test_inst_fns=None,
out_fns=None,
log_fns=None,
clas_dir=None,
descriptor=None,
timbl=None,
options="",
n=None,
log=False):
"""
Classify instance using Timbl in a cross-validation procedure.
@param inst_fns: a list of instance filenames for training; if no
test_inst_fns is supplied, the same files will be used for testing,
otherwise they are used for training only
@keyword test_inst_fns: a list of instance filenames for testing; this
allows for down-sampling of the training instances without affecting the
test instances
@keyword out_fns: list of classifier output files to be created
@keyword log_fns: list of classifier log files to be created;
ignored if keyword log is false
@keyword clas_dir: directory for creating classifier output files; ignored
if out_fns is given
@keyword descriptor: a Descriptor instance, required to infer the feature
metrics for Timbl, unless a TimblFile is supplied; ignored if timbl is
supplied
@keyword timbl: a tailored TimblFile instance; notice that it must at
least set the verbosity options +vo, +vdb, +vdi, and the -m option to
specify that the administrative features must be ignored.
@keyword options: list of additional Timbl options, excluding -f, -m, +vo,
+vdb, +vdi
@keyword n: limit on the number of cross-validations performed (by default
equals the number of instance filenames)
@keyword log: log Timbl's standard output and error streams to file
@return: list of Timbl output filenames
"""
if clas_dir:
makedirs(clas_dir)
if not timbl:
assert descriptor
timbl = TimblFile(default_opts=timbl_options_string(descriptor))
else:
# ignore descriptor
assert isinstance(timbl, TimblFile)
assert "+vo" in timbl.default_opts
assert "+vdb" in timbl.default_opts
assert "+vdi" in timbl.default_opts
assert "-m" in timbl.default_opts
return timbl.cross_validate(inst_fns,
test_inst_fns=test_inst_fns,
out_fns=out_fns,
log_fns=log_fns,
options=options,
n=n,
log=log,
out_dir=clas_dir)
