为了完成这个医学词向量,花了几天时间找了各种医学生物语料库和相关数据集并对其整理。整体语料库包括医学文献,医患对话,维基百科百度知道等医学相关语料,整体语料库共计1.6G,总共7052948句子,仅为生物医学领域相关语料。
使用专业医学类词汇进行分词(词汇表详见http://thuocl.thunlp.org/)
[('额叶', 0.4515002965927124), ('颞叶', 0.4498691260814667), ('枕叶', 0.38755619525909424), ('顶叶', 0.386254221200943), ('基底节', 0.381935179233551), ('岛叶', 0.35826876759529114), ('苍白球', 0.33769935369491577), ('尾状核', 0.33755943179130554), ('大脑半球', 0.33359262347221375), ('额页', 0.32096001505851746)]
[('头孢地尼', 0.5654973387718201), ('阿莫西林', 0.5394408106803894), ('头孢地尼胶囊', 0.5379139184951782), ('妇乐片', 0.5260443091392517), ('头孢地尼分散片', 0.5213251709938049), ('康妇炎胶囊', 0.5203120708465576), ('裸花紫珠胶囊', 0.5182883143424988), ('头孢克洛缓释片', 0.5178096294403076), ('头胞克洛', 0.5159974098205566), ('罗红霉素', 0.5115748643875122)]
[('海马牌', 0.6078361868858337), ('海马齿', 0.5532827377319336), ('普力马', 0.5418268442153931), ('马自达', 0.5407805442810059), ('东南汽车', 0.5387718677520752), ('000572', 0.5375587344169617), ('宝骏', 0.5361850261688232), ('海马回', 0.5352568030357361), ('北汽', 0.5325318574905396), ('小海马', 0.5315144062042236)]
[('头孢拉定', 0.7558029294013977), ('头孢菌素', 0.7490127086639404), ('头孢类', 0.7476578950881958), ('头孢氨苄', 0.7415952682495117), ('头孢曲松钠', 0.7406224608421326), ('头孢哌酮', 0.7398018836975098), ('头孢噻肟钠', 0.7393568158149719), ('头孢噻吩', 0.7348008751869202), ('头孢哌酮钠', 0.729317843914032), ('氨苄', 0.7292327284812927)]
model = word2vec.Word2Vec(sent, sg=0, epochs=8,vector_size=512, window=5, min_count=4, negative=3, sample=0.001, hs=1, workers=16)
想要使用只需要通过 gensim.models.KeyedVectors加载使用即可。
model = KeyedVectors.load_word2vec_format('Medical.txt', binary=False)