def calculate_sense_similarity(word, wiki_tag_file, keydict_path, similarity_function='jaccard', threshold=0):
tag_map, link_map = get_wiki_tag_and_link_maps(wiki_tag_file)
sense_idx_map = get_sense_idx_map(keydict_path, word)
for sense1, idx1 in sense_idx_map.iteritems():
tags1 = set(tag_map[sense1])
for sense2, idx2 in sense_idx_map.iteritems():
if sense1 != sense2:
tags2 = set(tag_map[sense2])
sim_score = None
if similarity_function == 'jaccard':
sim_score = calculate_jaccard_sim(tags1, tags2)
if sim_score > threshold:
print "{}\t{}-{}\t{}".format(word, sense1, sense2, sim_score)
def run():
model_id = sys.argv[1]
tag_map, link_map = get_wiki_tag_and_link_maps(sys.argv[2])
keydict_path = sys.argv[3]
output_fn = sys.argv[4]
words = sys.argv[5:]
model = load_model(model_id)
number_of_field = 15
table = [[] for _ in xrange(number_of_field)]
for word in words:
sense_idx_map = get_sense_idx_map(keydict_path, word)
i = 1
for sense, idx in sense_idx_map.iteritems():
try:
similar_words = model.most_similar(positive=[idx], topn=10)
table[i].extend([u"%s %s" % (t[0], t[1]) for t in similar_words])
table[i].append(u",".join(tag_map[sense]))
similarities = [t[1] for t in similar_words]
avg_sim = sum(similarities) / len(similarities)
table[i].append(u"%s %f" % (link_map[sense], avg_sim))
i += 1
except KeyError:
pass
for j in xrange(i, number_of_field):
diff = len(table[1]) - len(table[j])
for _ in xrange(diff):
table[j].append("")
for i in xrange(12):
if i == 6:
table[0].append(word)
else:
table[0].append("")
for column in table:
column.append(" ")
table_filtered = []
for column in table:
if len(column) != 0:
table_filtered.append(column)
headers = ["target_word"]
headers.extend(["sense-%d" % i for i in xrange(1, number_of_field)])
with codecs.open(output_fn, "w", "utf-8") as f:
t = tabulate.tabulate(zip(*table_filtered), tablefmt="simple", headers=headers)
f.write(t)
def run():
model_id = sys.argv[1]
tag_map, link_map = get_wiki_tag_and_link_maps(sys.argv[2])
keydict_path = sys.argv[3]
output_fn = sys.argv[4]
words = sys.argv[5:]
model = load_model(model_id)
number_of_field = 15
table = [[] for _ in xrange(number_of_field)]
for word in words:
sense_idx_map = get_sense_idx_map(keydict_path, word)
i = 1
for sense, idx in sense_idx_map.iteritems():
try:
similar_words = model.most_similar(positive=[idx], topn=10)
table[i].extend([u"%s %s" % (t[0], t[1]) for t in similar_words])
table[i].append(u",".join(tag_map[sense]))
similarities = [t[1] for t in similar_words]
avg_sim = sum(similarities) / len(similarities)
table[i].append(u"%s %f" % (link_map[sense], avg_sim))
i += 1
except KeyError:
pass
for j in xrange(i, number_of_field):
diff = len(table[1]) - len(table[j])
for _ in xrange(diff):
table[j].append("")
for i in xrange(12):
if i == 6:
table[0].append(word)
else:
table[0].append("")
for column in table:
column.append(" ")
table_filtered = []
for column in table:
if len(column) != 0:
table_filtered.append(column)
headers = ["target_word"]
headers.extend(["sense-%d" % i for i in xrange(1, number_of_field)])
with codecs.open(output_fn, 'w', 'utf-8') as f:
t = tabulate.tabulate(zip(*table_filtered), tablefmt="simple", headers=headers)
f.write(t)
def calculate_sense_similarity(word,
wiki_tag_file,
keydict_path,
similarity_function='jaccard',
threshold=0):
tag_map, link_map = get_wiki_tag_and_link_maps(wiki_tag_file)
sense_idx_map = get_sense_idx_map(keydict_path, word)
for sense1, idx1 in sense_idx_map.iteritems():
tags1 = set(tag_map[sense1])
for sense2, idx2 in sense_idx_map.iteritems():
if sense1 != sense2:
tags2 = set(tag_map[sense2])
sim_score = None
if similarity_function == 'jaccard':
sim_score = calculate_jaccard_sim(tags1, tags2)
if sim_score > threshold:
print "{}\t{}-{}\t{}".format(word, sense1, sense2,
sim_score)
