for label in labels:
ref = refsets[label]
test = testsets[label]
if not args.no_precision:
print '%s precision: %f' % (label, precision(ref, test) or 0)
if not args.no_recall:
print '%s recall: %f' % (label, recall(ref, test) or 0)
if not args.no_fmeasure:
print '%s f-measure: %f' % (label, f_measure(ref, test) or 0)
if args.show_most_informative and hasattr(classifier, 'show_most_informative_features') and not (args.multi and args.binary) and not args.cross_fold:
print '%d most informative features' % args.show_most_informative
classifier.show_most_informative_features(args.show_most_informative)
##############
## pickling ##
##############
if not args.no_pickle and not args.cross_fold:
if args.filename:
fname = os.path.expanduser(args.filename)
else:
name = '%s_%s.pickle' % (args.corpus, '_'.join(args.classifier))
fname = os.path.join(os.path.expanduser('~/nltk_data/classifiers'), name)
dump_object(classifier, fname, trace=args.trace)
################
## evaluation ##
################
if not args.no_eval:
if args.trace:
print "evaluating %s" % chunker.__class__.__name__
print chunker.evaluate(test_chunks)
##############
## pickling ##
##############
if not args.no_pickle:
if args.filename:
fname = os.path.expanduser(args.filename)
else:
# use the last part of the corpus name/path as the prefix
parts = [os.path.split(args.corpus.rstrip("/"))[-1]]
if args.classifier:
parts.append(args.classifier)
elif args.sequential:
parts.append(args.sequential)
name = "%s.pickle" % "_".join(parts)
fname = os.path.join(os.path.expanduser("~/nltk_data/chunkers"), name)
dump_object(chunker, fname, trace=args.trace)
##########################
## Hierarchical combine ##
##########################
labels = combined.labels()
label_classifiers = {}
for h in args.hierarchy:
label, path = h.split(':')
if label not in labels:
raise ValueError('%s is not in root labels: %s' % (label, labels))
label_classifiers[label] = nltk.data.load(path)
if args.trace:
print 'mapping %s to %s from %s' % (label, label_classifiers[label], path)
if label_classifiers:
if args.trace:
'combining %d label classifiers for root %s' % (len(label_classifiers), combined)
combined = multi.HierarchicalClassifier(combined, label_classifiers)
##############################
## dump combined classifier ##
##############################
fname = os.path.expanduser(args.filename)
dump_object(combined, fname, trace=args.trace)
################
if not args.no_eval:
print 'evaluating %s' % tagger.__class__.__name__
print 'accuracy: %f' % tagger.evaluate(test_sents)
##############
## pickling ##
##############
if not args.no_pickle:
if args.filename:
fname = os.path.expanduser(args.filename)
else:
# use the last part of the corpus name/path as the prefix
parts = [os.path.split(args.corpus.rstrip('/'))[-1]]
if args.brill:
parts.append('brill')
if args.classifier:
parts.append('_'.join(args.classifier))
if args.sequential:
parts.append(args.sequential)
name = '%s.pickle' % '_'.join(parts)
fname = os.path.join(os.path.expanduser('~/nltk_data/taggers'), name)
dump_object(tagger, fname, trace=args.trace)
