word2vec实践(一):预备知识
2016-06-15 00:00
316 查看
word2vec是google最新发布的深度学习工具,它利用神经网络将单词映射到低维连续实数空间,又称为单词嵌入。词与词之间的语义相似度可以通过两个单词的嵌入向量之间的余弦夹角直接衡量,更不用说使用诸如kmeans、层次聚类这样的算法来挖掘其功能了,同时作者TomasMikolov发现了比较有趣的现象,就是单词经过分布式表示后,向量之间依旧保持一定的语法规则,比如简单的加减法规则。
目前网络上有大量的实践文章和理论分析文章。主要列举如下:
理论分析文章:DeepLearning实战之word2vec
实践部分:
利用中文数据跑Google开源项目word2vec 分词工具ANSJ( 实例) Word2vec在事件挖掘中的调研参考文献:
[1]TomasMikolov,KaiChen,GregCorrado,andJeffreyDean.EfficientEstimationofWordRepresentationsinVectorSpace.InProceedingsofWorkshopatICLR,2013.[2]TomasMikolov,IlyaSutskever,KaiChen,GregCorrado,andJeffreyDean.DistributedRepresentationsofWordsandPhrasesandtheirCompositionality.InProceedingsofNIPS,2013.[3]TomasMikolov,Wen-tauYih,andGeoffreyZweig.LinguisticRegularitiesinContinuousSpaceWordRepresentations.InProceedingsofNAACLHLT,2013.[4]TomasMikolov,StefanKombrink,LukasBurget,JanCernocky,andSanjeevKhudanpur.Extensionsofrecurrentneuralnetworklanguagemodel.InAcoustics,SpeechandSignalProcessing(ICASSP),2011,IEEEInternationalConferenceon,pages5528–5531.IEEE,2011.[5]TomasMikolov,KaiChen,
3ff0
GregCorrado,andJeffreyDean.Efficientestimationofwordrepresentationsinvectorspace.ICLRWorkshop,2013.[6]FredericMorinandYoshuaBengio.Hierarchicalprobabilisticneuralnetworklanguagemodel.InProceedingsoftheinternationalworkshoponartificialintelligenceandstatistics,pages246–252,2005.[7]AndriyMnihandGeoffreyEHinton.Ascalablehierarchicaldistributedlanguagemodel.Advancesinneuralinformationprocessingsystems,21:1081–1088,2009.[8]Hinton,GeoffreyE."Learningdistributedrepresentationsofconcepts."Proceedingsoftheeighthannualconferenceofthecognitivesciencesociety.1986.[9]R.Rosenfeld,"Twodecadesofstatisticallanguagemodeling:wheredowegofromhere?",ProceedingsoftheIEEE,88(8),1270-1288,2000.[10]JeffreyDean,GregS.Corrado,RajatMonga,KaiChen,MatthieuDevin,QuocV.Le,MarkZ.Mao,Marc’AurelioRanzato,AndrewSenior,PaulTucker,KeYang,andAndrewY.Ng."LargeScaleDistributedDeepNetworks".ProceedingsofNIPS,2012.
[11]http://licstar.net/archives/328
[12]http://www.cs.columbia.edu/~mcollins/loglinear.pdf
[13]A.MnihandG.Hinton.Threenewgraphicalmodelsforstatisticallanguagemodelling.Proceedingsofthe24thinternationalconferenceonMachinelearning,pages641–648,2007[14]FredericMorinandYoshuaBengio.Hierarchicalprobabilisticneuralnetworklanguagemodel.InRobertG.CowellandZoubinGhahramani,editors,AISTATS’05,
相关文章推荐
- 词向量和语言模型
- Java内存泄露与溢出的区别
- [EN] TensorFlow Examples
- MapReduce:详解Shuffle过程
- 自然语言处理中的Attention Model:是什么及为什么
- myeclipse+maven实现多模块项目struts+spring+mybatis
- 短文本聚类方法
- 聚类算法-canopy
- 使用万能框架HttpHelper抓取安卓APP数据
- DBA应该掌握的SQL语句(三)
- 小白Windows7/10 64Bit安装Theano并实现GPU加速(没有MinGw等,详细步骤)
- 手把手入门神经网络系列(2)_74行代码实现手写数字识别
- ubuntu系统下eclipse配置hadoop开发环境并运行wordcount程序
- 泰迪杯比赛总结--关于NLP的资源
- Theano学习笔记(三)——图结构
- bash下快速移动光标的快捷键
- Python if 和 for 的多种写法
- 牛人推荐机器学习网站
- 关于Ubuntu12源列表的更改-source list
- 通过filter来改变request编码