def info(self):
"""
Print the feature statistics for the given model. (Assumes MaxEnt)
"""
mallet = c['mallet']
env = set_env_lang_utf8()
info_bin = os.path.join(os.path.join(mallet, 'bin'), 'classifier2info')
info_p = sub.Popen([info_bin, '--classifier', self._model],
stdout=sub.PIPE, stdin=sub.PIPE, stderr=sub.PIPE, env=env)
cur_class = None
feats = TwoLevelCountDict()
# Go through and pick out what the features are for
for line in info_p.stdout:
content = line.decode(encoding='utf-8')
class_change = re.search('FEATURES FOR CLASS (.*)', content)
# Set the current class if the section changes
if class_change:
cur_class = class_change.group(1).strip()
continue
# Otherwise, let's catalog the features.
word, prob = content.split()
feats.add(cur_class, word, float(prob))
# Now, print some info
for cur_class in feats.keys():
print(cur_class, end='\t')
print('%s:%.4f' % ('<default>', feats[cur_class]['<default>']), end='\t')
top_10 = feats.top_n(cur_class, n=10, key2_re='^nom')
print('\t'.join(['%s:%.4f' % (w,p) for w,p in top_10]))
def get_prototypes(tagged_path, proto_out, delimeter, ignoretags=[], unambiguous=False, maxproto=0):
encoding = getencoding(tagged_path)
tagged_file = codecs.open(tagged_path, "r", encoding=encoding)
tag_word_dict = TwoLevelCountDict()
word_tag_dict = TwoLevelCountDict()
proto_dict = defaultdict(set)
for line in tagged_file:
tokens = line.split()
for token in tokens:
word, pos = re.search("(^.*)%s(.*?)$" % delimeter, token).groups()
if pos not in ignoretags:
word = word.lower()
tag_word_dict.add(pos, word)
word_tag_dict.add(word, pos)
numproto = 0
# First, let's pick the maxproto most frequent words for a tag.
for tag in tag_word_dict.keys():
words = tag_word_dict[tag].most_frequent(minimum=1, num=None)
found_words = 0
for word in words:
freq_tag = word_tag_dict[word].most_frequent(minimum=1)
if freq_tag and freq_tag[0] == tag:
# if freq_tag:
proto_dict.add(freq_tag[0], word)
numproto += 1
found_words += 1
if maxproto and found_words == maxproto:
break
print("%s Prototypes found." % numproto)
# Now, set up the proto file for writing.
proto_file = open(proto_out, "w")
for tag in proto_dict:
proto_file.write(tag)
for word in proto_dict[tag]:
proto_file.write("\t" + word.lower()) # LOWERCASE for testing
proto_file.write("\n")
proto_file.close()
def _process_file(f):
c = TwoLevelCountDict()
d = TwoLevelCountDict()
m = TwoLevelCountDict()
print("Processing file {}".format(f))
xc = xc_load(f)
for inst in xc:
LOG.info("Now on instance {}".format(inst.id))
# Search for the gloss POS tier, if it exists.
gpos = inst.find(alignment=GLOSS_WORD_ID, type=POS_TIER_TYPE)
# If a gloss POS tier was found...
if gpos:
# Iterate through the projected tags.
for gp in gpos:
word = gp.igt.find(id=gp.attributes[ALIGNMENT])
grams = tokenize_item(word, morpheme_tokenizer)
# Add the (gram, POSTag) pair as something that was encountered.
for gram in grams:
m.add(gram.content.lower(), gp.value())
c.add(gp.value(), word.value().lower())
d.add(word.value().lower(), gp.value())
return (c, d, m)