def train(self): for dirPartition in self.dirPartitions: print "dirPartition", dirPartition (annotateDir, testDir, trainDir, modelDir, resultDir) = self.partitions.getDirTestNames(dirPartition) self._del_tmp_file(trainDir) # tmp file of test data are here bilbo = Bilbo(modelDir, self.bilboOptions, "crf_model_simple") # tmpFiles saved in modelDir if -k all bilbo.train(trainDir, modelDir, 1)
def train(self): for dirPartition in self.dirPartitions: print "dirPartition", dirPartition (annotateDir, testDir, trainDir, modelDir, resultDir) = self.partitions.getDirTestNames(dirPartition) #self._del_tmp_file(modelDir) bilbo = Bilbo(modelDir, self.bilboOptions, "crf_model_simple") # To save tmpFiles in modelDir bilbo.train(trainDir, modelDir, 1)
print "\t input data folder where the data files are (training or labeling)"
print " arg2 : <string>"
print "\t output data folder where the result files are saved\n"
else:
if options.g == "simple":
bilbo = Bilbo(str(args[1]), options, "crf_model_simple")
elif options.g == "detail":
bilbo = Bilbo(str(args[1]), options, "crf_model_detail")
dtype = options.t
if dtype == "bibl":
typeCorpus = 1
elif dtype == "note":
typeCorpus = 2
dirModel = os.path.join(rootDir, "model/corpus") + str(typeCorpus) + "/" + options.m + "/"
if not os.path.exists(dirModel):
os.makedirs(dirModel)
if options.T: # training
bilbo.train(str(args[0]), dirModel, typeCorpus)
elif options.L: # labeling
if dtype == "note" and options.e:
bilbo.annotate(str(args[0]), dirModel, typeCorpus, 1)
else:
bilbo.annotate(str(args[0]), dirModel, typeCorpus)
else:
print "Please choose training(-T option) or labeling(-L option)"
# simpleLabeling("Y.-M. KIM et al., An Extension of PLSA for Document Clustering, In Proceedings of ACM 17th Conference on Information and Knowledge Management, 2008.")
