svd.R

# 原始文件读入
txt <- readLines('txtdm.txt')
ignore = ",|:|!|'"
stopwords = c('and','edition','for','in','little','of','the','to')
txt <- tolower(txt)

# 文档分词
doc <- strsplit(txt,' ')
# 去除常用词和标点
doc <- lapply(doc,function(x)gsub(ignore,'',x))
doc <- lapply(doc,function(x){
  x[!(x %in% stopwords)]    
})

# 取词项集合
words <- unique(unlist(doc))

# 计算词项文档矩阵
DTM <- function(x,y){
  n <- length(x)
  m <- length(y)
  t <- matrix(nrow=n,ncol=m)
  for (i in 1:n){
    for (j in 1:m){
     t[i,j] <- sum(doc[[j]]==words[i])
    }
  }
  return(t)
}
t <- DTM(words,doc)

# 只取同时出现在两个以上文档中的词项
DocsPerWord <- rowSums(t>0) 
words <- words[DocsPerWord>1]
t <- DTM(words,doc)

# 将频数转为tfidf值
TFIDF <- function(t){
  WordsPerDoc <- colSums(t)     
  DocsPerWord <- rowSums(t>0) 
  for (i in 1:nrow(t)){
    for (j in 1:ncol(t)){
      t[i,j] <- (t[i,j]/WordsPerDoc[j])*log(ncol(t)/DocsPerWord[i])
    }
  }
  return(t)
}
tfidf <- TFIDF(t)

# SVD分解
res <- svd(tfidf)
# 词项语义相关矩阵
datau <- data.frame(res$u[,2:3])
# 文档语义相关矩阵
datav <- data.frame(res$v[,2:3])

library(ggplot2)
p <- ggplot()+
  geom_point(data=datau,aes(X1,X2))+
  geom_point(data=datav,aes(X1,X2),
             size=3,color='red4')+
  geom_text(data=datau,aes(X1,X2),
            label=words,vjust=2)+
  geom_text(data=datav,aes(X1,X2),
            label=1:9,vjust=2)+
  theme_bw()
print(p)