精华内容
下载资源
问答
  • 课程分享——人工智能商业实战应用:金融知识图谱构建实战【企业内训现场实录】视频教程,完整版,附源码课件。欢迎大家下载学习。
  • 分享一套视频教程,名字叫——人工智能商业实战应用:金融知识图谱构建实战【企业内训现场实录】,非常棒的一套金融知识图谱课程,附源码+课件资料,完整版,值得学习
  • 知识图谱构建实战PPT
  • 知识图谱实战构建红楼梦知识图谱

    万次阅读 多人点赞 2020-09-14 20:04:12
    视频内容简介,最后可以带大家确实做一个知识图谱 b站课程地址:https://www.bilibili.com/video/BV11k4y1y7A3 补充内容: 允许neo4j数据库后,输入如下地址即可进入数据库:http://localhost:7474/browser/ py2...

    本文为数据集整理以及代码存放,本内容已经录制b站课程,如有需要可以前去观看,建议点赞投币~

    视频内容简介,最后可以带大家确实做一个知识图谱

    b站课程地址:https://www.bilibili.com/video/BV11k4y1y7A3

    补充内容:

    允许neo4j数据库后,输入如下地址即可进入数据库:http://localhost:7474/browser/

    py2neo版本为:4.1

    效果图:

    基础代码:

    from py2neo import Graph,Node,Relationship,NodeMatcher
    import py2neo
    #py2neo、import、neo2j自带的语言
    #账号密码改为自己的即可
    g=Graph('http://localhost:7474',user='neo4j',password='neo4j')
    #g.run('match (n) detach delete n’)
    test_node_1=Node("Person",name="b站one")
    test_node_2=Node("Person",name="b站two")
    test_node_1['age']=18
    test_node_1['sex']='男'
    test_node_2['age']=16
    test_node_2['sex']='女'
    #创建结点
    #g.create(test_node_1)
    #g.create(test_node_2)
    #覆盖式创建结点
    g.merge(test_node_1,"Person","name")
    g.merge(test_node_2,"Person","name")
    friend=Relationship(test_node_1,'friend',test_node_2)
    g.merge(friend,"Person","name")
    matcher=NodeMatcher(g)
    print(matcher.match("Person",name="b站one").first())

    创建知识图谱代码:

    import csv
    import py2neo
    from py2neo import Graph,Node,Relationship,NodeMatcher
    #账号密码改为自己的即可
    g=Graph('http://localhost:7474',user='neo4j',password='neo4j')
    with open('/Users/ren/Desktop/triples.csv','r',encoding='utf-8') as f:
        reader=csv.reader(f)
        for item in reader:
            if reader.line_num==1:
                continue
            print("当前行数:",reader.line_num,"当前内容:",item)
            start_node=Node("Person",name=item[0])
            end_node=Node("Person",name=item[1])
            relation=Relationship(start_node,item[3],end_node)
            g.merge(start_node,"Person","name")
            g.merge(end_node,"Person","name")
            g.merge(relation,"Person","name")
    #以下为neo4j代码,如需代码运行,请放入g.run(...)内运行,将...替换为下列代码
    MATCH (p: Person {name:"贾宝玉"})-[k:丫鬟]-(r)
    return p,k,r
    MATCH (p1:Person {name:"贾宝玉"}),(p2:Person{name:"香菱"}),p=shortestpath((p1)-[*..10]-(p2))
    RETURN p

    由于分享百度网盘链接容易挂,这里直接把数据集放这里了,复制到文本中改成.csv文件即可

    triples.csv:(数据集来源:http://www.openkg.cn/home

    "head","tail","relation","label"
    "贾代善","贾源","son","子"
    "娄氏","贾源","daughter_in_law_of_grandson","重孙媳妇"
    "贾母","贾代善","wife","妻"
    "老姨奶奶","贾代善","concubine","妾"
    "贾敏","贾代善","daughter","女"
    "嫣红","贾赦","concubine","妾"
    "翠云","贾赦","concubine","妾"
    "娇红","贾赦","concubine","妾"
    "贾迎春","贾赦","daughter","女"
    "赵姨娘","贾政","concubine","妾"
    "周姨娘","贾政","concubine","妾"
    "贾珠","贾政","son","子"
    "尤二姐","贾琏","concubine","妾"
    "秋桐","贾琏","concubine","妾"
    "平儿","贾琏","concubine","妾"
    "薛宝衩","贾宝玉","wife","妻"
    "花袭人","贾宝玉","concubine","妾"
    "贾桂","贾宝玉","son","子"
    "贾菌","娄氏","son","子"
    "周秀才","贾巧姐","brother_in_law","姐夫"
    "周财主","周秀才","father","父亲"
    "周妈妈","周秀才","mother","母亲"
    "孙亲太太","孙绍祖","mother","母亲"
    "刑大舅二姐","邢夫人","younger_sister","妹"
    "邢德全","邢夫人","younger_brother","弟"
    "张大老爷","邢夫人","old_relatives","老亲"
    "邢忠","邢夫人","elder_brother","兄"
    "张大老爷之女","张大老爷","daughter","女"
    "刑秞烟","邢忠","daughter","女"
    "李婶","李守中","sister_in_law","弟媳"
    "李婶之弟","李婶","younger_brother","弟"
    "李纹","李婶","daughter","女"
    "甄宝玉","李绮","husband","夫"
    "甄应嘉","甄宝玉祖母","son","子"
    "甄夫人","甄应嘉","wife","妻"
    "大姑娘","甄应嘉","daughter","女"
    "二姑娘","甄应嘉","daughter","女"
    "三姑娘","甄应嘉","daughter","女"
    "袭人之母","袭人","mother","母亲"
    "花自芳","袭人","elder_brother","兄"
    "周琼之子","贾探春","husband","夫"
    "林如海之父","林如海","father","父亲"
    "林如海之祖","林如海","grandfather","祖父"
    "林如海之子","林如海","son","子"
    "李妈","贾巧姐","nurser","乳母"
    "昭儿","贾琏","servant","奴仆"
    "兴儿","贾琏","servant","奴仆"
    "隆儿","贾琏","servant","奴仆"
    "庆儿","贾琏","servant","奴仆"
    "赵嬷嬷","贾琏","nurser","乳母"
    "赵天梁","赵嬷嬷","son","子"
    "赵天栋","赵嬷嬷","son","子"
    "王信","贾琏","servant","奴仆"
    "王信媳妇","贾琏","servant","奴仆"
    "鲍二","贾琏","servant","奴仆"
    "善姐","尤二姐","servant_girl","丫鬟"
    "迎春乳母","贾迎春","nurser","乳母"
    "王住儿媳妇","迎春乳母","daughter_in_law","子媳"
    "绣桔","贾迎春","servant_girl","丫鬟"
    "莲花儿","贾迎春","servant_girl","丫鬟"
    "秦思祺","贾迎春","servant_girl","丫鬟"
    "潘又安","秦思祺","boy_friend","男友"
    "贾琮奶妈","贾琮","nurser","乳母"
    "篆儿","刑秞烟","servant_girl","丫鬟"
    "新进来的奶子","贾兰","nurser","奶妈"
    "素云","李纨","servant_girl","丫鬟"
    "碧月","李纨","servant_girl","丫鬟"
    "抱琴","贾元春","servant_girl","丫鬟"
    "李嬷嬷","贾宝玉","nurser","乳母"
    "赵嬷嬷","贾宝玉","nurser","乳母"
    "张嬷嬷","贾宝玉","nurser","乳母"
    "宋嬷嬷","贾宝玉","servant","奴仆"
    "叶茗烟","贾宝玉","servant","奴仆"
    "锄药","贾宝玉","servant","奴仆"
    "扫红","贾宝玉","servant","奴仆"
    "墨雨","贾宝玉","servant","奴仆"
    "引泉","贾宝玉","servant","奴仆"
    "扫花","贾宝玉","servant","奴仆"
    "挑云","贾宝玉","servant","奴仆"
    "伴鹤","贾宝玉","servant","奴仆"
    "双瑞","贾宝玉","servant","奴仆"
    "双寿","贾宝玉","servant","奴仆"
    "老叶妈","叶茗烟","mother","母亲"
    "李贵","李嬷嬷","son","子"
    "李嬷嬷孙子","贾宝玉","servant","奴仆"
    "王荣","贾宝玉","servant","奴仆"
    "张若锦","贾宝玉","servant","奴仆"
    "赵亦华","贾宝玉","servant","奴仆"
    "周瑞","王夫人","servant","奴仆"
    "周瑞女儿","周瑞","daughter","女"
    "冷子兴","周瑞","son_in_law","女婿"
    "周嫂子的儿子","周瑞","son","子"
    "何三","周瑞","adopted_son","干儿子"
    "白金钏","王夫人","servant_girl","丫鬟"
    "白玉钏","王夫人","servant_girl","丫鬟"
    "彩云","王夫人","servant_girl","丫鬟"
    "彩鸾","王夫人","servant_girl","丫鬟"
    "绣鸾","王夫人","servant_girl","丫鬟"
    "绣凤","王夫人","servant_girl","丫鬟"
    "白老媳妇","白金钏","mother","母亲"
    "白老媳妇","白玉钏","mother","母亲"
    "彩霞之母","彩霞","mother","母亲"
    "小霞","彩霞","younger_sister","妹"
    "小鹊","赵姨娘","servant_girl","丫鬟"
    "小吉祥","赵姨娘","servant_girl","丫鬟"
    "赵国基","赵姨娘","elder_brother","兄"
    "钱槐","赵姨娘","elder_male_cousin","表兄"
    "钱槐","贾环","servant","奴仆"
    "雪雁","林黛玉","servant_girl","丫鬟"
    "王嬷嬷","林黛玉","nurser","乳母"
    "紫鹊","林黛玉","servant_girl","丫鬟"
    "春纤","林黛玉","servant_girl","丫鬟"
    "藕官","林黛玉","servant_girl","丫鬟"
    "赖嬷嬷","赖大","mother","母亲"
    "赖尚荣","赖大","son","子"
    "赖大的女儿","赖大","daughter","女"
    "来旺儿","王熙凤","servant","奴仆"
    "旺儿媳妇","王熙凤","servant_girl","丫鬟"
    "来喜家的","王熙凤","servant","奴仆"
    "丰儿","王熙凤","servant_girl","丫鬟"
    "彩明","王熙凤","servant_girl","丫鬟"
    "林红玉","王熙凤","servant_girl","丫鬟"
    "林之孝","林红玉","father","父亲"
    "王善保","邢夫人","servant","奴仆"
    "费婆子","邢夫人","servant_girl","丫鬟"
    "金鸳鸯","贾母","servant_girl","丫鬟"
    "琥珀","贾母","servant_girl","丫鬟"
    "鹦鹉","贾母","servant_girl","丫鬟"
    "珍珠","贾母","servant_girl","丫鬟"
    "翡翠","贾母","servant_girl","丫鬟"
    "玻璃","贾母","servant_girl","丫鬟"
    "文官","贾母","servant_girl","丫鬟"
    "傻大姐的娘","傻大姐","mother","母亲"
    "金彩","鸳鸯","father","父亲"
    "金文翔","鸳鸯","elder_brother","兄"
    "多官","晴雯","elder_male_cousin","表哥"
    "多浑虫父亲","晴雯","mothers_brother","舅"
    "多姑娘儿","多官","wife","妻"
    "吴贵","晴雯","elder_male_cousin","表兄"
    "花芳官","贾宝玉","servant_girl","丫鬟"
    "媚人","贾宝玉","servant_girl","丫鬟"
    "麝月","贾宝玉","servant_girl","丫鬟"
    "茜雪","贾宝玉","servant_girl","丫鬟"
    "秋纹","贾宝玉","servant_girl","丫鬟"
    "绮霞","贾宝玉","servant_girl","丫鬟"
    "碧痕","贾宝玉","servant_girl","丫鬟"
    "檀云","贾宝玉","servant_girl","丫鬟"
    "四儿","贾宝玉","servant_girl","丫鬟"
    "佳蕙","贾宝玉","servant_girl","丫鬟"
    "坠儿","贾宝玉","servant_girl","丫鬟"
    "紫绡","贾宝玉","servant_girl","丫鬟"
    "良儿","贾宝玉","servant_girl","丫鬟"
    "何春燕","贾宝玉","servant_girl","丫鬟"
    "厨房中的柳家媳妇","柳五儿","mother","母亲"
    "柳二媳妇的妹子","柳五儿","younger_sister","妹"
    "哥嫂侄儿","柳五儿","nephew","侄"
    "侍书","贾探春","servant_girl","丫鬟"
    "翠墨","贾探春","servant_girl","丫鬟"
    "艾官","贾探春","servant_girl","丫鬟"
    "小蝉","贾探春","servant_girl","丫鬟"
    "何婆","何春燕","mother","母亲"
    "何婆","芳官","adopted_mother","干娘"
    "小鸠儿","何春燕","younger_sister","妹"
    "夏婆子","何春燕","maternal_aunt","姨妈"
    "夏婆子","藕官","adopted_mother","干娘"
    "夏婆子","小蝉","grandmother","外祖母"
    "蒋玉菡","贾宝玉","friend","朋友"
    "通判傅试","贾政","pupil","门生"
    "傅秋芳","傅试","younger_sister","妹"
    "太祖皇帝","先皇","father","父亲"
    "太上皇","先皇","son","子"
    "皇太后","太上皇","wife","妻"
    "太妃","太上皇","kings_concubine","嫔妃"
    "皇帝","太上皇","son","子"
    "贾元春","皇帝","imperial_concubine","妃"
    "吴贵妃","皇帝","imperial_concubine","妃"
    "周贵人","皇帝","imperial_concubine","妃"
    "周贵人父亲","周贵人","father","父亲"
    "吴天佑","吴贵妃","father","父亲"
    "戴权","皇帝","minister","臣"
    "夏守忠","皇帝","minister","臣"
    "贾代化","贾演","son","子"
    "焦大","贾演","servant","奴仆"
    "贾敷","贾代化","son","子"
    "贾敬","贾代化","son","子"
    "贾惜春","贾敬","daughter","女"
    "史湘云爷爷","史侯","son","子"
    "史鼐","史湘云爷爷","son","子"
    "史鼎","史湘云爷爷","son","子"
    "湘云母","史湘云爷爷","daughter_in_law","儿媳"
    "史鼎的夫人","史鼎","wife","妻"
    "史湘云","湘云母","daughter","女"
    "卫若兰","史湘云","husband","夫"
    "翠缕","史湘云","servant_girl","丫鬟"
    "葵官","史湘云","servant_girl","丫鬟"
    "周奶妈","史湘云","nurser","乳母"
    "凤姐之祖王夫人之父","王公","son","子"
    "王夫人之大兄凤姐之父","凤姐之祖王夫人之父","son","子"
    "王子腾","凤姐之祖王夫人之父","son","子"
    "王子胜","凤姐之祖王夫人之父","son","子"
    "王夫人","凤姐之祖王夫人之父","daughter","女"
    "薛姨妈","凤姐之祖王夫人之父","daughter","女"
    "王成父","凤姐之祖王夫人之父","nephew","侄"
    "老舅太太","王夫人之大兄凤姐之父","wife","妻"
    "王仁","王夫人之大兄凤姐之父","son","子"
    "王熙凤","王夫人之大兄凤姐之父","daughter","女"
    "王子腾夫人","王子腾","wife","妻"
    "王子腾之女","王子腾","daughter","女"
    "保宁侯之子","王子腾之女","husband","夫"
    "王成","王成父","son","子"
    "王狗儿","王成","son","子"
    "刘氏","王狗儿","wife","妻"
    "王青儿","王狗儿","daughter","女"
    "王板儿","王狗儿","son","子"
    "刘姥姥","刘氏","mother","母亲"
    "宝钗祖父","薛公","son","子"
    "薛公之孙","宝钗祖父","son","子"
    "薛宝琴父","宝钗祖父","son","子"
    "薛姨妈","薛公之孙","wife","妻"
    "薛宝钗","薛公之孙","daughter","女"
    "薛蟠","薛公之孙","son","子"
    "同喜","薛姨妈","servant_girl","丫鬟"
    "同贵","薛姨妈","servant_girl","丫鬟"
    "莺儿","薛宝钗","servant_girl","丫鬟"
    "文杏","薛宝钗","servant_girl","丫鬟"
    "喜儿","薛宝钗","servant_girl","丫鬟"
    "蕊官","薛宝钗","servant_girl","丫鬟"
    "夏金桂","薛蟠","wife","妻"
    "老苍头","薛蟠","meme_ama","乳父"
    "宝蟾","夏金桂","servant_girl","丫鬟"
    "小舍儿","夏金桂","servant_girl","丫鬟"
    "夏奶奶","夏金桂","mother","母亲"
    "夏三","夏奶奶","son","子"
    "臻儿","香菱","servant_girl","丫鬟"
    "薛宝琴母","薛宝琴父","wife","妻"
    "薛蝌","薛宝琴父","son","子"
    "薛宝琴","薛宝琴父","daughter","女"
    "邢岫烟","薛蝌","wife","妻"
    "梅翰林之子","薛宝琴","husband","夫"
    "小螺","薛宝琴","servant_girl","丫鬟"
    "荳官","薛宝琴","servant_girl","丫鬟"
    "梅翰林","梅翰林之子","father","父亲"
    "贾瑞之父","贾代儒","son","子"
    "贾瑞之母","贾瑞之父","wife","妻"
    "贾瑞","贾瑞之父","son","子"
    "贾琼","贾琼之母","son","子"
    "贾四姐","贾琼之母","daughter","女"
    "贾王扁","贾王扁之母","son","子"
    "贾喜鸾","贾王扁之母","daughter","女"
    "贾芸","五嫂子卜氏","son","子"
    "卜世仁","五嫂子卜氏","brother","兄弟"
    "小丫头子","贾芸","servant_girl","丫鬟"
    "倪儿娘子","倪二","wife","妻"
    "倪二女儿","倪二","daughter","女"
    "马贩子王短腿","倪二","friend","友人"
    "卜世仁娘子","卜世仁","wife","妻"
    "卜银姐","卜世仁","daughter","女"
    "王奶奶","卜世仁","neighbour","邻居"
    "卜世仁店里伙计","卜世仁","partner","伙计"
    "贾芹","周氏","son","子"
    "金氏","贾璜","wife","妻"
    "胡氏","金氏","elder_sister_in_law","嫂"
    "金荣","胡氏","son","子"
    "娇杏","贾化","wife","妻"
    "贾雨村子","贾化","son","子"
    "应天府门子","贾化","son","子"
    "张如圭","贾化","friend","友人"
    "王老爷","贾化","friend","友人"
    "天子之妻","应天府门子","wife","妻"
    "东安郡王穆莳拜","东平郡王","son","子"
    "南安王太妃","南安郡王","wife","妻"
    "南安郡王之孙","南安郡王","grandson","孙子"
    "西宁郡王之孙","西宁郡王","grandson","孙子"
    "西宁郡王妃","西宁郡王","wife","妻"
    "北静王太妃","北静郡王","daughter_in_law","儿媳"
    "北静王少妃","水溶","wife","妻"
    "北静郡王长府官","水溶","minister","臣"
    "镇国公诰命","牛清","wife","妻"
    "镇国公诰命长男","牛清","son","子"
    "牛继宗","牛清","grandson","孙子"
    "柳芳","柳彪","grandson","孙子"
    "陈瑞文","陈翼","grandson","孙子"
    "马尚","马魁","grandson","孙子"
    "侯孝康","侯明","grandson","孙子"
    "缮国公诰命","缮国公","wife","妻"
    "石光珠","缮国公","grandson","孙子"
    "蒋子宁","平原侯","grandson","孙子"
    "谢鲸","定城侯","grandson","孙子"
    "戚建辉","襄阳侯","grandson","孙子"
    "裘良","景田侯","grandson","孙子"
    "锦乡侯诰命","锦乡侯","wife","妻"
    "韩奇","锦乡侯","son","子"
    "冯紫英","神武将军冯唐","son","子"
    "锦田侯诰命","马道婆","donor","施主"
    "蕊官","地藏庵的圆心","pupil","徒弟"
    "藕官","圆心","pupil","徒弟"
    "智能儿","净虚","pupil","徒弟"
    "智善","净虚","pupil","徒弟"
    "智通","净虚","pupil","徒弟"
    "于老爷","净虚","donor","施主"
    "胡老爷","净虚","donor","施主"
    "胡老爷太太","净虚","donor","施主"
    "胡老爷公子","净虚","donor","施主"
    "张大财主","净虚","donor","施主"
    "花芳官","智能","pupil","徒弟"
    "张金哥","张大财主","daughter","女"
    "李公子","张金哥","employer","雇主"
    "长安原任守备","李公子","father","父亲"
    "李衙内","长安府府太爷","mothers_brother","小舅子"
    "贾代化","贾演","son","子"
    "焦大","贾演","servant","奴仆"
    "贾敷","贾代化","son","子"
    "贾敬","贾代化","son","子"
    "尤氏","尤老娘","daughter","女"
    "贾蔷","贾演","great_great_grandson","玄孙"
    "龄官","贾蔷","girl_friend","女友"
    "佩凤","贾珍","concubine","妾"
    "偕鸾","贾珍","concubine","妾"
    "文花","贾珍","concubine","妾"
    "茄官","尤氏","servant_girl","丫鬟"
    "银蝶","尤氏","servant_girl","丫鬟"
    "炒豆儿","尤氏","servant_girl","丫鬟"
    "秦可卿","贾蓉","wife","妻"
    "胡氏","贾蓉","step_wife","续弦"
    "许氏","贾蓉","step_wife","续弦"
    "入画","贾惜春","servant_girl","丫鬟"
    "彩屏","贾惜春","servant_girl","丫鬟"
    "彩儿","贾惜春","servant_girl","丫鬟"
    "瑞珠","秦可卿","servant_girl","丫鬟"
    "宝珠","秦可卿","servant_girl","丫鬟"
    "秦钟","秦业","son","子"
    "智能儿","秦钟","girl_friend","女友"
    "净虚","智能儿","master","师父"
    "智善","净虚","pupil","徒弟"
    "智通","净虚","pupil","徒弟"
    "喜儿","贾珍","servant","奴仆"
    "寿儿","贾珍","servant","奴仆"
    "尤三姐","柳湘莲","girl_friend","女友"
    "杏奴","柳湘莲","servant","奴仆"
    "尤二姐","贾琏","concubine","妾"
    "善姐","尤二姐","servant_girl","丫鬟"
    "甄宝玉","李绮","husband","夫"
    "贾演","贾源","elder_brother","兄"
    "贾赦","贾代善","son","子"
    "贾政","贾代善","son","子"
    "贾敏","贾政","younger_sister","妹"
    "邢夫人","贾赦","wife","妻"
    "贾琏","贾赦","son","子"
    "贾琮","贾赦","son","子"
    "王夫人","贾政","wife","妻"
    "贾元春","贾政","daughter","女"
    "贾宝玉","贾政","son","子"
    "贾探春","贾政","daughter","女"
    "贾环","贾政","son","子"
    "王熙凤","贾琏","wife","妻"
    "贾巧姐","贾琏","daughter","女"
    "贾兰","贾珠","son","子"
    "李纨","贾珠","wife","妻"
    "孙绍祖","贾迎春","husband","夫"
    "李守中","李纨","father","父亲"
    "李绮","李婶","daughter","女"
    "甄宝玉","甄应嘉","son","子"
    "林如海","贾敏","husband","夫"
    "林黛玉","林如海","daughter","女"
    "王嬷嬷","贾宝玉","nurser","乳母"
    "彩霞","王夫人","servant_girl","丫鬟"
    "傻大姐","贾母","servant_girl","丫鬟"
    "晴雯","贾宝玉","servant_girl","丫鬟"
    "篆儿","贾宝玉","servant_girl","丫鬟"
    "柳五儿","贾宝玉","servant_girl","丫鬟"
    "贾珍","贾敬","son","子"
    "贾母","史侯","daughter","女"
    "香菱","薛蟠","concubine","妾"
    "倪二","贾芸","friend","友人"
    "水溶","北静郡王","grandson","孙子"
    "尤氏","贾珍","wife","妻"
    "尤二姐","尤老娘","daughter","女"
    "尤三姐","尤老娘","daughter","女"
    "贾蓉","贾珍","son","子"
    "秦可卿","秦业","daughter","女"
    

     

    展开全文
  • 项目实战:如何构建知识图谱

    万次阅读 2018-05-10 14:26:15
    实践了下怎么建一个简单的知识图谱,两个版本,一个从 0 开始(start from scratch),一个在 CN-DBpedia 基础上补充,把 MySQL,PostgreSQL,Neo4j 数据库都尝试了下。自己跌跌撞撞摸索可能踩坑了都不知道,欢迎讨论...

    实践了下怎么建一个简单的知识图谱,两个版本,一个从 0 开始(start from scratch),一个在 CN-DBpedia 基础上补充,把 MySQL,PostgreSQL,Neo4j 数据库都尝试了下。自己跌跌撞撞摸索可能踩坑了都不知道,欢迎讨论。

    CN-DBpedia 构建流程

    知识库可以分为两种类型,一种是以 Freebase,Yago2 为代表的 Curated KBs,主要从维基百科和 WordNet 等知识库中抽取大量的实体及实体关系,像是一种结构化的维基百科。另一种是以 Stanford OpenIE,和我们学校 Never-Ending Language Learning (NELL) 为代表的 Extracted KBs,直接从上亿个非结构化网页中抽取实体关系三元组。与 Freebase 相比,这样得到的知识更加多样性,但同时精确度要低于 Curated KBs,因为实体关系和实体更多的是自然语言的形式,如“奥巴马出生在火奴鲁鲁。” 可以被表示为(“Obama”, “was also born in”, “ Honolulu”),

    下面以 CN-DBpedia 为例看下知识图谱大致是怎么构建的。
    CN_DBpedia%E6%9E%84%E5%BB%BA%E6%B5%81%E7%A8%8B.png
    CN_DBpedia.png

    上图分别是 CN-DBpedia 的构建流程和系统架构。知识图谱的构建是一个浩大的工程,从大方面来讲,分为知识获取、知识融合、知识验证、知识计算和应用几个部分,也就是上面架构图从下往上走的一个流程,简单来走一下这个流程。

    数据支持层

    最底下是知识获取及存储,或者说是数据支持层,首先从不同来源、不同结构的数据中获取知识,CN-DBpedia 的知识来源主要是通过爬取各种百科知识这类半结构化数据。

    至于数据存储,要考虑的是选什么样的数据库以及怎么设计 schema。选关系数据库还是NoSQL 数据库?要不要用内存数据库?要不要用图数据库?这些都需要根据数据场景慎重选择。CN-DBpedia 实际上是基于 mongo 数据库,参与开发的谢晨昊提到,一般只有在基于特定领域才可能会用到图数据库,就知识图谱而言,基于 json(bson) 的 mongo 就足够了。用到图查询的领域如征信,一般是需要要找两个公司之间的关联交易,会用到最短路径/社区计算等。

    schema 的重要性不用多说,高质量、标准化的 schema 能有效降低领域数据之间对接的成本。我们希望达到的效果是,对于任何数据,进入知识图谱后后续流程都是相同的。换言之,对于不同格式、不同来源、不同内容的数据,在接入知识图谱时都会按照预定义的 schema 对数据进行转换和清洗,无缝使用已有元数据和资源。

    知识融合层

    我们知道,目前分布在互联网上的知识常常以分散、异构、自治的形式存在,另外还具有冗余、噪音、不确定、非完备的特点,清洗并不能解决这些问题,因此从这些知识出发,通常需要融合验证的步骤,来将不同源不同结构的数据融合成统一的知识图谱,以保证知识的一致性。所以数据支持层往上一层实际上是融合层,主要工作是对获取的数据进行标注、抽取,得到大量的三元组,并对这些三元组进行融合,去冗余、去冲突、规范化,

    第一部分 SPO 三元组抽取,对不同种类的数据用不同的技术提取

    • 从结构化数据库中获取知识:D2R
      难点:复杂表数据的处理
    • 从链接数据中获取知识:图映射
      难点:数据对齐
    • 从半结构化(网站)数据中获取知识:使用包装器
      难点:方便的包装器定义方法,包装器自动生成、更新与维护
    • 从文本中获取知识:信息抽取
      难点:结果的准确率与覆盖率
    %E7%9F%A5%E8%AF%86%E8%8E%B7%E5%8F%96.png

    尤其是纯文本数据会涉及到的 实体识别、实体链接、实体关系识别、概念抽取 等,需要用到许多自然语言处理的技术,包括但不仅限于分词、词性标注、分布式语义表达、篇章潜在主题分析、同义词构建、语义解析、依存句法、语义角色标注、语义相似度计算等等。

    第二部分才到融合,目的是将不同数据源获取的知识进行融合构建数据之间的关联。包括 实体对齐、属性对齐、冲突消解、规范化 等,这一部分很多都是 dirty work,更多的是做一个数据的映射、实体的匹配,可能还会涉及的是本体的构建和融合。最后融合而成的知识库存入上一部分提到的数据库中。如有必要,也需要如 Spark 等大数据平台提供高性能计算能力,支持快速运算。

    知识融合的四个难点:

    • 实现不同来源、不同形态数据的融合
    • 海量数据的高效融合
    • 新增知识的实时融合
    • 多语言的融合

    知识验证

    再往上一层主要是验证,分为补全、纠错、外链、更新各部分,确保知识图谱的一致性和准确性。一个典型问题是,知识图谱的构建不是一个静态的过程,当引入新知识时,需要判断新知识是否正确,与已有知识是否一致,如果新知识与旧知识间有冲突,那么要判断是原有的知识错了,还是新的知识不靠谱?这里可以用到的证据可以是权威度、冗余度、多样性、一致性等。如果新知识是正确的,那么要进行相关实体和关系的更新。

    知识计算和应用

    这一部分主要是基于知识图谱计算功能以及知识图谱的应用。知识计算主要是根据图谱提供的信息得到更多隐含的知识,像是通过本体或者规则推理技术可以获取数据中存在的隐含知识;通过链接预测预测实体间隐含的关系;通过社区计算在知识网络上计算获取知识图谱上存在的社区,提供知识间关联的路径……通过知识计算知识图谱可以产生大量的智能应用如专家系统、推荐系统、语义搜索、问答等。

    知识图谱涉及到的技术非常多,每一项技术都需要专门去研究,而且已经有很多的研究成果。Anyway 这章不是来论述知识图谱的具体技术,而是讲怎么做一个 hello world 式的行业知识图谱。这里讲两个小 demo,一个是 爬虫+mysql+d3 的小型知识图谱,另一个是 基于 CN-DBpedia+爬虫+PostgreSQL+d3 的”增量型”知识图谱,要实现的是某行业上市公司与其高管之间的关系图谱。

    Start from scratch

    数据获取

    第一个重要问题是,我们需要什么样的知识?需要爬什么样的数据?一般在数据获取之前会先做个知识建模,建立知识图谱的数据模式,可以采用两种方法:一种是自顶向下的方法,专家手工编辑形成数据模式;另一种是自底向上的方法,基于行业现有的标准进行转换或者从现有的高质量行业数据源中进行映射。数据建模都过程很重要,因为标准化的 schema 能有效降低领域数据之间对接的成本。

    作为一个简单的 demo,我们只做上市公司和高管之间的关系图谱,企业信息就用公司注册的基本信息,高管信息就用基本的姓名、出生年、性别、学历这些。然后开始写爬虫,爬虫看着简单,实际有很多的技巧,怎么做优先级调度,怎么并行,怎么屏蔽规避,怎么在遵守互联网协议的基础上最大化爬取的效率,有很多小的 trick,之前博客里也说了很多,就不展开了,要注意的一点是,高质量的数据来源是成功的一半!

    来扯一扯爬取建议:

    • 从数据质量来看,优先考虑权威的、稳定的、数据格式规整且前后一致、数据完整的网页
    • 从爬取成本来看,优先考虑免登录、免验证码、无访问限制的页面
    • 爬下来的数据务必保存好爬取时间、爬取来源(source)或网页地址(url)
      source 可以是新浪财经这类的简单标识,url 则是网页地址,这些在后续数据清洗以及之后的纠错(权威度计算)、外链和更新中非常重要

    企业信息可以在天眼查、启信宝、企查查各种网站查到,信息还蛮全的,不过有访问限制,需要注册登录,还有验证码的环节,当然可以过五关斩六将爬到我们要的数据,然而没这个必要,换别个网站就好。

    推荐两个数据来源:

    其中巨潮资讯还可以同时爬取高管以及公告信息。看一下数据
    cfi.png
    cninfo.png

    换句话说,我们直接能得到规范的实体(公司、人),以及规范的关系(高管),当然也可以把高管展开,用下一层关系,董事长、监事之类,这就需要做进一步的清洗,也可能需要做关系的对齐。

    这里爬虫框架我用的是 scrapy+redis 分布式,每天可以定时爬取,爬下来的数据写好自动化清洗脚本,定时入库。

    数据存储

    数据存储是非常重要的一环,第一个问题是选什么数据库,这里作为 starter,用的是关系型数据库 MySQL。设计了四张表,两张实体表分别存公司(company)人物(person)的信息,一张关系表存公司和高管的对应关系(management),最后一张 SPO 表存三元组

    为什么爬下来两张表,存储却要用 4 张表?
    一个考虑是知识图谱里典型的一词多义问题,相同实体名但有可能指向不同的意义,比如说 Paris 既可以表示巴黎,也可以表示人名,怎么办?让作为地名的 “Paris” 和作为人的 “Paris” 有各自独一无二的ID。“Paris1”(巴黎)通过一种内在关系与埃菲尔铁塔相联,而 “Paris2”(人)通过取消关系与各种真人秀相联。这里也是一样的场景,同名同姓不同人,需要用 id 做唯一性标识,也就是说我们需要对原来的数据格式做一个转换,不同的张三要标识成张三1,张三2… 那么,用什么来区别人呢?拍脑袋想用姓名、生日、性别来定义一个人,也就是说我们需要一张人物表,需要 (name, birth, sex) 来作composite unique key 表示每个人。公司也是相同的道理,不过这里只有上市公司,股票代码就可以作为唯一性标识。

    Person 表和 company 表是多对多的关系,这里需要做 normalization,用 management 这张表来把多对多转化为两个一对多的关系,(person_id, company_id) 就表示了这种映射。management 和 spo 表都表示了这种映射,为什么用两张表呢?是出于实体对齐的考虑。management 保存了原始的关系,”董事”、监事”等,而 spo 把这些关系都映射成”高管”,也就是说 management 可能需要通过映射才能得到 SPO 表,SPO 才是最终成型的表。

    可能有更简单的方法来处理上述问题,思考中,待更新—-

    我们知道知识库里的关系其实有两种,一种是属性(property),一种是关系(relation)。那么还有一个问题是 SPO 需不需要存储属性?
    spo.png

    最后要注意的一点是,每条记录要保存创建时间以及最后更新时间,做一个简单的版本控制。

    数据可视化

    Flask 做 server,d3 做可视化,可以检索公司名/人名获取相应的图谱,如下图。之后会试着更新有向图版本。
    eg.png

    Start from CN-DBpedia

    把 CN-DBpedia 的三元组数据,大概 6500 万条,导入数据库,这里尝试了 PostgreSQL。然后检索了 112 家上市公司的注册公司名称,只有 69 家公司返回了结果,属性、关系都不是很完善,说明了通用知识图谱有其不完整性(也有可能需要先做一次 mention2entity,可能它的标准实体并不是注册信息的公司名称,不过 API 小范围试了下很多是 Unknown Mention)。

    做法也很简单,把前面 Start from scratch 中得到的 SPO 表插入到这里的 SPO 表就好了。这么简单?因为这个场景下不用做实体对齐和关系对齐。

    拓展

    这只是个 hello world 项目,在此基础上可以进行很多有趣的拓展,最相近的比如说加入企业和股东的关系,可以进行企业最终控制人查询(e.g.,基于股权投资关系寻找持股比例最大的股东,最终追溯至自然人或国有资产管理部门)。再往后可以做企业社交图谱查询、企业与企业的路径发现、企业风险评估、反欺诈等等等等。具体来说:

    1. 重新设计数据模型 引入”概念”,形成可动态变化的“概念—实体—属性—关系”数据模型,实现各类数据的统一建模
    2. 扩展多源、异构数据,结合实体抽取、关系抽取等技术,填充数据模型
    3. 展开知识融合(实体链接、关系链接、冲突消解等)、验证工作(纠错、更新等)

    最后补充一下用 Neo4j 方式产生的可视化图,有两种方法。一是把上面说到的 MySQL/PostgreSQL 里的 company 表和 person 表存成 node,node 之间的关系由 spo 表中 type == relation 的 record 中产生;二是更直接的,从 spo 表中,遇到 type == property 就给 node(subject) 增加属性({predicate:object}),遇到 type == relation 就给 node 增加关系((Nsubject) - [r:predicate]-> node(Nobject)),得到下面的图,移动鼠标到相应位置就可以在下方查看到关系和节点的属性。

    show2.png

    项目地址

    http://www.shuang0420.com/2017/09/05/%E9%A1%B9%E7%9B%AE%E5%AE%9E%E6%88%98-%E7%9F%A5%E8%AF%86%E5%9B%BE%E8%B0%B1%E5%88%9D%E6%8E%A2/

    展开全文
  • 行业知识图谱概述,包括行业图谱简介,行业知识图谱的应用及挑战,以及行业知识图谱生命周期管理。...行业知识图谱应用实战,以金融证券行业应用为例,演示知识图谱从知识建模、知识抽取到行业应用的全过程。
  • 知识图谱_完整_项目实战_(附源码)的代码。文件tree:kgcar.zip,splider.zip,Videolink.txt。赠送没有密码的视频。保证真实有效。出现问题可私信解决。
  • 人工智能商业实战应用:金融知识图谱构建实战【企业内训现场实录】,完整版,附源码。 此课程的背景是一个金融知识图谱的大项目,用于构建A股公司的知识图谱,并基于知识图谱提供语义搜索、智能问答等服务。 课程...
  • python+Neo4j+flask,汽车行业知识图谱项目实战视频+源码(不加密).txt
  • 当关系型数据库oracle、mysql或者hive中存在一张关于某个主题的表时,我们应该如何基于该表创建知识图谱? 我们来看一个简单的例子。 01 关系型表 这张表结构如上图所示,包含公众号名称,建立时间,传播知识的主题...

    在这里插入图片描述
    请关注一下微信公众号:机器学习简明教程

    当关系型数据库oracle、mysql或者hive中存在一张关于某个主题的表时,我们应该如何基于该表创建知识图谱?

    我们来看一个简单的例子。

    01 关系型表

    在这里插入图片描述
    这张表结构如上图所示,包含公众号名称,建立时间,传播知识的主题,作者共4个字段。

    节点的创建有两种方法。第一种方法,可以把每个字段都做成节点,公众号名称字段就是节点的标签名,具体的字段值就是name属性值;第二种方法,可以把公众号名称做成一种节点,其他字段信息做成公众号信息节点。

    关系的创建其实就是把上面这张表列转行,关系的标签就是字段名:建立于、传播和作者。

    详细工程代码和描述如下。

    02 抽取

    from py2neo import Node, Graph, Relationship
    import pandas as pd
    # connect neo4j
    graph = Graph("bolt://localhost:7687", username="neo4j", password="****")
    label_1 = "公众号节点"
    label_2 = "公众号信息节点"
    graph.delete_all()
    ​
    # step1:read data
    data = pd.read_csv("./data/raw_data.csv", header=0)
    data["建立于"] = data["建立于"].astype(str)
    ​
    # step2 : extract nodes
    node_list = list(set(data['公众号名']))
    

    mac下执行命令pip install py2neo==3,安装指定版本的py2neo。

    实例化Graph类,并通过该类连接neo4j。

    第一类节点:抽取出“公众号名称”字段,去重后转成list。

    # step3 : extract nodes
    node_info_list = []
    for i in list(data.columns)[1:]:
        node_info_list.extend(data[i])
    node_info_list = list(set(node_info_list))
    

    第二类节点:遍历除了“公众号名称”之外的所有字段值,去重后,放进node_info_list中。

    # step4 : extract relationships
    relation_data = pd.DataFrame()
    for i in list(data.columns)[1:]:
        rel_data = data[["公众号名", i]]
        rel_data["关系"] = i
        rel_data.columns = ["公众号节点", "公众号信息节点", "关系"]
        relation_data = pd.concat([relation_data, rel_data], axis=0)
    

    关系三元组:基于原始的data,列转行抽取关系。

    03 写入

    def create_node(node_list, label):
    for name in node_list:
            print(name)
            name_node = Node(label, name=name)
            print(name_node)
            graph.create(name_node)
    create_node(node_list, label_1)
    create_node(node_info_list, label_2)
    

    创建节点:create_node遍历节点列表,创建标签为label的节点。

    def create_relation(relation_data, label_a, label_b):
    for m in range(0, len(relation_data)):
            print()
            rel = Relationship(
                graph.find_one(label_a, property_key="name", property_value=str(list(relation_data['公众号节点'])[m])),
                list(relation_data['关系'])[m],
                graph.find_one(label_b, property_key="name", property_value=str(list(relation_data['公众号信息节点'])[m])))
            graph.merge(rel, label=[label_b, label_a])
     create_relation(relation_data, label_1, label_2)
    

    创建关系:create_relation遍历关系数据并创建关系。
    在这里插入图片描述

    日志:节点和关系抽取过程如上。

    在这里插入图片描述
    知识图谱:最后neo4j中会出现如上图所示的节点与关系。

    展开全文
  • 知识图谱实战系列(Python版)视频教程,2020年录制,配套有源码、课件和数据集,完整版;知识图谱实战系列课程旨在帮助同学们快速...通俗讲解核心技术点及其应用领域,全程实战演示如何构建知识图谱生态中各项核心技术。
  • OpenKG地址:http://openkg.cn/tool/gbuilder网站地址:http://gbuilder.gstore.cn知识图谱能够让机器去理解和认知世界中的事物和现象,...

    OpenKG地址:http://openkg.cn/tool/gbuilder

    网站地址:http://gbuilder.gstore.cn


    知识图谱能够让机器去理解和认知世界中的事物和现象,并解释现象出现的原因,推理出隐藏在数据之间深层的、隐含的关系,使得知识图谱技术从最初谷歌用来提升搜索引擎的结果来增强用户体验,到现在已经被金融、公安、能源、教育、医疗等领域众多行业进行大量运用。

    2fdb10c9e5ad42c07959acb69910e33a.png

    知识图谱作为大数据和人工智能时代的关键技术已经让越来越多的人意识到它的重要性和价值。知识图谱的应用现在处于“百花齐放”的状态。

    知识图谱全生命周期分为构建、存储管理、应用三个阶段。

    32e00b1f0ee7b5f6b479fcc829f3ed23.png

    知识图谱的应用已经受到业内广泛关注,知识图谱的存储管理也有众多解决方案。例如我们前期研发的开源知识图谱图数据库系统gStore(http://www.gstore.cn/pcsite/index.html#/)就是知识图谱存储的工具。gStore在OpenKG上也有介绍 (http://www.openkg.cn/tool/gstore )

    然而知识图谱构建却鲜有统一化的平台工具,但是这是知识图谱生命周期的技术难点之一。

    244026f664c55bd6e5f59e7a1d021b55.png

    这是由于知识图谱构建是一项需要花费大量的人力和时间,却不直接体现价值的工作,但知识图谱构建却是最基础、最关键的工作,是解决“巧妇难为无米之炊”窘境的核心手段。

    c7b0232073beb4529ce4d838caaa7f3e.png

    北京大学王选计算机研究所和大数据分析与应用技术国家工程实验室(北京大学)邹磊教授团队通过两年时间,打造了知识图谱自动化构建平台gBuilder。gBuilder基于机器学习、自然语言处理、图数据库等技术可以实现对结构化数据和非结构化数据的知识抽取,并转化为知识图谱三元组。

    28212590584af841d11a783345b27b16.png

    Schema设计

    无论是结构化项目还是非结构化项目,均需首先设计知识图谱Schema。知识图谱Schema一方面可以描述知识图谱中的实体类型、实体属性和关系等信息,另一方面也是知识图谱查询和分析的重要参考,相当于关系型数据库的表结构。

    同现有的Schema设计方法不一样的是,gBuilder的Schema设计模块是一个轻量级的Web平台,以图的方式来表述知识图谱Schema,用户可以通过拖拽的方式在画布上设计类、类属性和关系。

    2659e61434301c08c5550424e2b4bd64.gif

    结构化数据抽取

    对于结构化项目而言,其知识抽取流程设计就是将结构化表及字段,与Schema中的实体类型、属性、关系等进行映射,并形成映射文件。

    1b9f942eeae123e66e483159ae1032b7.png

    gBuilder自动化构建平台结构化数据抽取基于D2RQ平台,让用户显式地、可视化地处理结构化数据抽取的所有步骤,摆脱复杂的映射语言,易于使用。当前gBuilder能从MySQL、Oracle、SQL Server、PostgreSQL、达梦等关系型数据库中将数据100%准确的映射为RDF三元组数据。

    51bac716a0ca5f84ee18516d03eca791.gif

    非结构化数据抽取

    对于非结构化数据抽取而言是当前知识图谱图谱构建的重难点,例如给下图一段文字,如何将里面的实体以及实体之间的关系准确抽取出来是一个关键问题。当前业内对于非结构数据的自动抽取产品还是较为欠缺。

    cbba916f30eaabcbcbe39b47bd0e64c5.png

    对于非结构化项目而言,需要通过加载数据集、设计构建流程、开始构建、构建结束步骤。在构建流程中gBuilder通过可视化拖拽的方式来自定义构建非结构化数据知识图谱构建流程。同时gBuilder提供了大量的可选模型,它们有着不同的特性,以及在不同的数据上预训练,用户可以根据需求,挑选出最合适的模型作为构建模型,也可以使用自己的数据训练模型进行图谱构建。构建过程中可以随时使用系统自带的流程检查功能和测试功能来测试流程的正确性与构造效果。完成构建后,可以查看构建的结果,从非结构化数据集中抽取出了构建知识图谱所需要的三元组。

    45b09a5e860c6096903b24ff9cd7dfe4.png

    具体抽取流程示例如下,用户首先根据实际业务场景需要通过拖拽算子的方式构建数据抽取流程,然后输入非结构化数据,最后抽取出RDF三元组数据。例如输入“小明是小王的爸爸”测试数据,通过流程的一步步运行,最终抽取出“<小明> <父亲> <小王> ”三元组数据。

    86a47c0c934ea1b7a0f06d6c7507be77.gif

    在gBuilder平台的非结构化数据抽取详细操作如下面动图所示:

    aee11f7f3ca01d3bc89118fb20fba4e0.gif

    最后通过gBuilder抽取的RDF三元组数据同gBuilder团队研发的gStore图数据库系统无缝衔接,再加上该团队研发的面向知识图谱自然语言问答引擎gAnswer,形成了覆盖知识图谱构建、知识图谱存储管理和知识图谱应用的完整生命周期的知识图谱一体化解决方案。

    f538852d16f9decdc8009a3da174e9fc.png

    gBuilder整体采用微服务架构,主要模块包括可视化Flowline工具库,数据管理模块,项目管理模块、模型库和任务中心;其中模型库与任务中心等高负载组件采用多云融合方案,为用户提供弹性、实时和可扩展的知识图谱构建服务。

    ec511a726831d34dbb6cc052c056e6e1.png

    系统框架图

    gBuilder具有可视化、易于使用、高扩展、高延伸、灵活性以及支持多种语言(英语、中文)的特性,有力的支撑知识图谱的构建,打破“巧妇难为无米之炊”的窘境。后续gBuilder也将支持更多数据模型的知识图谱构建以丰富知识图谱构建生态和实际业务需要。也期待更多同行者加入我们一起打造和完善图谱生态。


    OpenKG

    OpenKG(中文开放知识图谱)旨在推动以中文为核心的知识图谱数据的开放、互联及众包,并促进知识图谱算法、工具及平台的开源开放。

    fd4b57752b90285599c691fbf5971514.png

    点击阅读原文,进入 OpenKG 网站。

    展开全文
  • 领域(学科)知识图谱构建心得

    千次阅读 2021-01-05 21:37:25
    1、清华大学杨玉基的“一种准确而高效的领域知识图谱构建方法”.讲述了怎么通过4步进行半自动话的构建领域知识图谱,参考价值极大,步骤清晰。一种准确而高效的领域知识图谱构建方法 - 中国期刊全文数据库 2、华东...
  • 知识图谱实战案例完全剖析(附完整源码和数据集)Python与Neo4j的集成
  • 给出了意图知识图谱的定义、构建框架和政府治理场景的使用范例,展示了意图知识图谱是解决政府治理的效果评估问题的一种有效方法。在政府治理的场景中,利用意图知识图谱可以分析对比同一治理话题下不同治理主体之间...
  • 知识图谱实战例子

    2021-10-19 21:10:49
    当时上知识图谱课程,整理的几个有关知识图谱实战例子。 1.知识图谱简介相关 这是一份通俗易懂的知识图谱技术与应用指南 2.知识图谱相关例子 基于 REfO 的 KBQA 实现及示例 基于Python模块RefO实现的知识库问答...
  • 知识图谱实战系列课程旨在帮助同学们快速掌握知识图谱领域核心技术,基于Python各大开源技术实现知识图谱核心应用。通俗讲解核心技术点及其应用领域,全程实战演示如何构建知识图谱生态中各项核心技术。
  • [python案例]金融知识图谱构建流程

    千次阅读 2020-04-24 08:30:00
    向AI转型的程序员都关注了这个号????????????机器学习AI算法工程 公众号:datayx小型金融知识图谱构流程示范存储方式基于RDF的存储基于图数据库的存储知识图谱构建代码...
  • 知识图谱实战

    2020-12-18 10:50:54
    原标题:知识图谱实战知识图谱是近来非常红火的技术,融合网络爬虫,自然语言处理,机器学习,深度学习,图数据库,复杂网络分析等多种热门技术于一身,技术含量密集,在构造语义搜索,问答平台,高智能的人机界面等...
  • 深度学习对比知识图谱能够实现端到端的模型,中间减少人为的参与,知识图谱通过三元组的关系表示,可以最大限度的获得自然世界中的相互联系 发展方向 Pretrain+finetune 预训练:大语料、无监督、深模型获得语义表示...
  • 昨天在北理工参加了一场由 雪晴数据网和北京理工大学大数据创新学习中心...1.佛学知识图谱构建技术东南大学 漆桂林教授1.1 什么是知识?1.2 知识图谱为搜索引擎带来的补充作用!1.3知识图谱的几个关键技术1.data extra
  • 本文转载自公众号:博文视点Broadview。 AI是新的生产力,知识图...
  • 从具体案例了解知识图谱构建流程

    万次阅读 多人点赞 2018-05-07 20:04:35
    经济责任审计知识图谱构建流程总结。     自2012年Google提出“知识图谱”的概念以来,知识图谱就一直是学术研究的重要方向,现在有很多高校、企业都致力于将这项技术应用到医疗、教育、商业等领域,并且已经...

空空如也

空空如也

1 2 3 4 5 ... 20
收藏数 3,759
精华内容 1,503
关键字:

知识图谱构建实战