from infomap import Infomap
import tqdm
min_sim = 0.5
word_vecs = np.array(temp_v)
#word_vecs = model.wv.vectors
word_vecs /= (word_vecs ** 2).sum(axis=1, keepdims=True) ** 0.5
word2id = {j: i for i, j in enumerate(sub_n_words)}
new_words = sub_n_words
new_vecs = word_vecs
links = {}
# 每个词找与它相似度不小于0.6的词(不超过50个),来作为图上的边
for i in range(len(new_words)):
sims = np.dot(new_vecs, new_vecs[i])
idxs = sims.argsort()[::-1][1:]
for j in idxs:
if sims[j] >= min_sim:
links[(i, j)] = float(sims[j])
else:
break
infomapWrapper = Infomap("--two-level --directed")
for (i, j), sim in links.items():
_ = infomapWrapper.addLink(i, j, sim)
infomapWrapper.run()
tree = infomapWrapper.tree
word2class = {}
class2word = {}
for node in tree:
if node.is_leaf:
if new_words[node.node_id] not in word2class:
word2class[new_words[node.node_id]] = []
word2class[new_words[node.node_id]].append(node.module_id)
if node.module_id not in class2word:
class2word[node.module_id] = []
class2word[node.module_id].append(new_words[node.node_id])
for i in class2word.keys():
print(class2word[i])
infomap代码实现
最新推荐文章于 2024-05-23 09:45:21 发布