docvec使用 相似文本

# 需要的包
from gensim.models import Doc2Vec
from gensim.models.doc2vec import TaggedDocument

# 处理数据
for i, doc in enumerate(docs):
    # document = TaggedDocument(words=doc.split(" "), tags=['TRAIN' + '_%s' % i])
    document = TaggedDocument(words=doc.split(" "), tags=[tags[i]])
    x_train.append(document)

# 训练

def train():
    x_train = load_data(); size = 100
    model_dm = Doc2Vec(x_train, min_count=1, window=3, size=size, sample=1e-3, negative=5, workers=4)
    model_dm.train(x_train, total_examples=model_dm.corpus_count, epochs=5)
    model_dm.save('./static/model/9.model')
    return model_dm

# 找相似文本

def text_sim(text):
    model_dm = Doc2Vec.load('../static/d2v_model/9.model')
    inferred_vector_dm = model_dm.infer_vector(text.split(' '))
    sims = model_dm.docvecs.most_similar([inferred_vector_dm], topn=200)
    return [sim[0] for sim in sims]

# tag和words字典

_ = dict((d.tags[0], d.words) for d in data)

可以用来调整相似文本推荐中的权重