csv <- read.table("body1.txt",header=T)

csv  <-t(csv)

mystopwords<-unlist (read.table("StopWords.txt",stringsAsFactors=F))

library(tm)

#移除数字

removeNumbers = function(x) { ret = gsub("[0-9０１２３４５６７８９]","",x) }

#中文分词，也可以考虑使用rmmseg4j、rsmartcn

wordsegment<- function(x) {

library(Rwordseg)

segmentCN(x)

}

#去除停止词，效果比较差，可以进一步完善

removeStopWords = function(x,words) {

ret = character(0)

index <- 1

it_max <- length(x)

while (index <= it_max) {

if (length(words[words==x[index]]) <1) ret <- c(ret,x[index])

index <- index +1

}

ret

}

Sys.setenv(JAVA_HOME='C:/ProgramFiles/Java/jdk1.6.0_43/jre')

sample.words <- lapply(csv, removeNumbers)

sample.words <- lapply(sample.words, wordsegment)

#先处理中文分词，再处理stopwords，防止全局替换丢失信息

sample.words <- lapply(sample.words, removeStopWords, mystopwords)

#构建语料库

corpus = Corpus(VectorSource(sample.words))

sample.dtm<- DocumentTermMatrix(corpus, control = list(wordLengths = c(2, Inf)))

dtm2 = removeSparseTerms(sample.dtm,sparse=0.99)

d <- dist(dtm2, method ="euclidean")

fit <- hclust(d,method="ward.D")

plot(fit)

cutree(fit,k=8)#聚类的图显示的是8类，但数据太多特别密集，也看不出有多少数据，不知道是1700条还是3500条