如何从R中的文档术语矩阵中删除空文档

Question

I am performing kmeans clustering for twitter data, for which I am cleaning the tweets and creating a corpus. Later I find the dtm and use the tf-idf theory.

But my dtm has few empty documents which I want to remove because kmeans can't run for empty docs.

Here is my code:

removeURL <- function(x) gsub("http[[:alnum:][:punct:]]*", "", x) 
replacePunctuation <- function(x)
{
  x <- tolower(x)
  x <- gsub("[.]+[ ]"," ",x)
  x <- gsub("[:]+[ ]"," ",x)
  x <- gsub("[?]"," ",x)
  x <- gsub("[!]"," ",x)
  x <- gsub("[;]"," ",x)
  x <- gsub("[,]"," ",x)
  x <- gsub("[@]"," ",x)
  x <- gsub("[???]"," ",x)
  x <- gsub("[€]"," ",x)
  x

}

myStopwords <- c(stopwords('english'), "rt")


#preprocessing
tweet_corpus <- Corpus(VectorSource(tweet_raw$text))
tweet_corpus_clean <- tweet_corpus %>%
  tm_map(content_transformer(tolower)) %>% 
  tm_map(removeNumbers) %>%
  tm_map(removeWords,myStopwords) %>%
  tm_map(content_transformer(replacePunctuation)) %>%
  tm_map(stripWhitespace)%>%
  tm_map(content_transformer(removeURL))


dtm <- DocumentTermMatrix(tweet_corpus_clean ) 

#tf-idf

mat4 <- weightTfIdf(dtm) #when i run this, i get 2 docs that are empty
mat4 <- as.matrix(mat4)  

Answer 1

Obviously you can't do that with another tm_map .

But the text mining package also has tm_filter , which you can use to filter empty documents.

Answer 2

If your document does not contain any entry/word, then you could do this:

rowSumDoc <- apply(dtm, 1, sum) 
dtm2 <- dtm[rowSumDoc > 0, ] 

Basically, above we are summing the words in each document first. Later, we are subsetting dtm for documents that are not empty based on earlier summation of words in each document.