R: inspect Document Term Matrix results in Error: Repeated indices currently not allowed

问题

I have the following dummy data:

final6 <- data.frame(docname = paste0("doc", 1:6),
                  articles = c("Catalonia independence in matter of days",
                               "Anger over Johnson Libya bodies comment",
                               "Man admits frenzied mum and son murder",
                               "The headache that changed my life",
                               "Las Vegas killer sick, demented - Trump",
                               "Instagram baby photo scammer banned")
)

And I want to create a DocumentTermMatrix with reference to document names (that I could later link to the original article text). To achieve this, I follow instruction from this post:

myReader <- readTabular(mapping=list(content="articles", id="docname"))
text_corpus <- VCorpus(DataframeSource(final6), readerControl = list(reader = myReader))

# define function that replaces ounctuation with spaces 
replacePunctuation <- content_transformer(function(x) {return (gsub("[[:punct:]]"," ", x))}) # replaces punctuation with empty spaces

# remove customised words 
myWords <- c("ok", "chat", 'okay', 'day', 'today', "might", "bye", "hello", "thank", "you", "please", "sorry", "hello", "hi")

# clean text 
cleantext <- function(corpus){
  clean_corpus <- tm_map(corpus, removeNumbers)
  clean_corpus <- tm_map(clean_corpus, tolower)
  clean_corpus <- tm_map(clean_corpus, PlainTextDocument)
  clean_corpus <- tm_map(clean_corpus, replacePunctuation)
  clean_corpus <- tm_map(clean_corpus, removePunctuation)
  clean_corpus <- tm_map(clean_corpus, removeWords, c(stopwords("english"), myWords, top_names))
  clean_corpus <- tm_map(clean_corpus, stripWhitespace)
  clean_corpus <- tm_map(clean_corpus, stemDocument, language = "english")

  clean_corpus
}

clean_corpus <- cleantext(text_corpus) 

# create dtm
chat_DTM <- DocumentTermMatrix(clean_corpus, control = list(wordLengths = c(3, Inf)))

Now, when I want to inspect the matrix, I get the error:

inspect(chat_DTM)

Error in [.simple_triplet_matrix(x, docs, terms) : Repeated indices currently not allowed.

To be fair, this error occurs even if I create a corpus based on text only and without passing doc id as an attribute. Any ideas what causes the problem?

回答1:

The problem was with the PlainTextDocument function that removes meta data from corpus. If you modify clean_text function as follows, this results in the clean DTM that can be inspected without any errors returned:

cleantext <- function(corpus){
  clean_corpus <- tm_map(corpus, removeNumbers)
  clean_corpus <- tm_map(clean_corpus, content_transformer(tolower)) #!! modified
  #clean_corpus <- tm_map(clean_corpus, PlainTextDocument) ### !!!! PlainTextDocument function erases metadata from corpus = document id! So this needs to be erased
  clean_corpus <- tm_map(clean_corpus, replacePunctuation)
  clean_corpus <- tm_map(clean_corpus, removePunctuation)
  clean_corpus <- tm_map(clean_corpus, removeWords, c(stopwords("english"), myWords, top_names))
  clean_corpus <- tm_map(clean_corpus, stripWhitespace)
  clean_corpus <- tm_map(clean_corpus, stemDocument, language = "english")

  clean_corpus
}

clean_corpus <- cleantext(text_corpus)


chat_DTM2 <- DocumentTermMatrix(clean_corpus)
inspect(chat_DTM2)

The answer was inspired by this solution. Thanks!

回答2:

You might get a similar error if creating a directory source by using DirSource(recursive=T, ...), and 2 or more files, in different paths, have the same name.

In this case, a workaround is:

ds   <- DirSource(".", recursive=T)
ovid <- VCorpus(ds)
names(ovid) <- ds$filelist

来源：https://stackoverflow.com/questions/47410866/r-inspect-document-term-matrix-results-in-error-repeated-indices-currently-not

标签

text

corpus