快速进行词向量训练和读取

1.词向量训练demo

from gensim.models import Word2Vec
from gensim.test.utils import common_texts
import jieba
import tqdm

word2vec_path = './resources/word2vec.model'

def word_vector_gener():
    """
    几种不同的方法来生成词向量
    :return:
    """
    # 1.word2vec
    # 获取原始数据
    DATA_PATH = './data/seo_search_word_copy.txt'
#     word2evctor = open('./word2vector.txt', 'w', encoding='utf8')
    word_list = []
    finall = []
    # jieba分词
    with open(DATA_PATH, 'r', encoding='utf8') as file:
        for each_line in tqdm.tqdm(file.readlines()):
            query = each_line.strip().split('	')[-1]
            # 分词
            cut_word = jieba.lcut(query)
            finall.append(cut_word)
    # 训练模型
    model = Word2Vec(finall, sg=1, size=10, window=2, min_count=1, negative=1,
                 sample=0.001, workers=4)
#     model.save('./resources/word2vec.model')
    model.wv.save(word2vec_path)
    print(model['老师'])


if __name__ == '__main__':
    word_vector_gener()

2.词向量加载demo(此方法为获得词向量最快)

word2vec_path = './resources/word2vec.model'
wv = KeyedVectors.load(word2vec_path, mmap='r')
vector = wv['主管'] 
word = wv.most_similar(['主管'], topn=30)

print(word)

输出：

[('组长', 0.8488447070121765),
 ('经理', 0.8272342085838318),
 ('总监', 0.816636323928833),
 ('副经理', 0.8071938753128052),
 ('部长', 0.8019827604293823),
 ('专员', 0.7792257070541382),
 ('高级专员', 0.7695066332817078),
 ('主任', 0.7676611542701721),
 ('负责人', 0.761403501033783),
 ('部副', 0.7570186853408813),
 ('及', 0.7355248928070068),
 ('业务主管', 0.732032299041748),
 ('岗', 0.7316986322402954),
 ('副总', 0.7278518676757812),
 ('科长', 0.72648024559021),
 ('兼', 0.7262977957725525),
 ('助理', 0.7255839705467224),
 ('资深', 0.7252861261367798),
 ('组', 0.7167786955833435),
 ('储干', 0.7150581479072571),
 ('班长', 0.7146369218826294),
 ('职员', 0.7104721665382385),
 ('实习生', 0.707991898059845),
 ('支持', 0.7070707082748413),
 ('高级', 0.7055947184562683),
 ('管理人员', 0.7054109573364258),
 ('初级', 0.7042156457901001),
 ('副理', 0.7038965821266174),
 ('小组长', 0.7035383582115173),
 ('技术主管', 0.7024495601654053)]