Loop through a tm corpus without losing corpus structure

Question

I have a tm corpus of documents and a list of words. I want to run a for loop over the corpus, so that the loop removes each word in the list from the corpus sequentially.

Some replication data:

library(tm)
m <- cbind(c("Apple blue two","Pear yellow five","Banana yellow two"),
             c(1, 2, 3))
tm_corpus <- Corpus(VectorSource(m[,1]))
words <- as.list(c("Apple", "yellow", "two"))

tm_corpus is now a corpus object consisting of 3 documents:

<<SimpleCorpus>>
Metadata:  corpus specific: 1, document level (indexed): 0
Content:  documents: 3

words is a list of 3 words:

[[1]]
[1] "Apple"

[[2]]
[1] "yellow"

[[3]]
[1] "two"

I have tried three different loops. The first one is:

tm_corpusClean <- tm_corpus
for (i in seq_along(tm_corpusClean)) {
  for (u in seq_along(words)) {
    tm_corpusClean[i] <- tm_map(tm_corpusClean[i], removeWords, words[[u]])
  }
}

Which returns the following error 7 times (numbered 1-7):

Error in x$dmeta[i, , drop = FALSE] : incorrect number of dimensions
In addition: Warning messages:
1: In tm_corpusClean[i] <- tm_map(tm_corpusClean[i], removeWords,                 
words[[u]]) :
  number of items to replace is not a multiple of replacement length
2: In tm_corpusClean[i] <- tm_map(tm_corpusClean[i], removeWords,         
words[[u]]) :
  number of items to replace is not a multiple of replacement length
[...]

The second one is:

tm_corpusClean <- tm_corpus
for (i in seq_along(words)) {
  for (u in seq_along(tm_corpusClean)) {
    tm_corpusClean[u] <- tm_map(tm_corpusClean[u], removeWords, words[[i]])
  }
}

Which returns the error:

Error in x$dmeta[i, , drop = FALSE] : incorrect number of dimensions

The last loop is:

tm_corpusClean <- tm_corpus
for (i in seq_along(words)) {
  tm_corpusClean <- tm_map(tm_corpusClean, removeWords, words[[i]])
}

This actually returns an object named tm_corpusClean, but this object only returns the first document instead of all original three:

inspect(tm_corpusClean[[1]])

<<PlainTextDocument>>
Metadata:  7
Content:  chars: 6

 blue

Where am I going wrong?

Adam Quek Adam Quek · Accepted Answer · 2017-04-25T08:36:33

Before we go to the sequential removal, test if tm_map work on your example:

obj1 <- tm_map(tm_corpus, removeWords, unlist(words))
sapply(obj1, `[`, "content")

$`1.content`
[1] " blue "

$`2.content`
[1] "Pear  five"

$`3.content`
[1] "Banana  "

Next, use lapply to sequentially remove one word at a time, i.e. "Apple", "yellow", "two":

obj2 <- lapply(words, function(word) tm_map(tm_corpus, removeWords, word))
sapply(obj2, function(x) sapply(x, `[`, "content"))

          [,1]                [,2]             [,3]              
1.content " blue two"         "Apple blue two" "Apple blue "     
2.content "Pear yellow five"  "Pear  five"     "Pear yellow five"
3.content "Banana yellow two" "Banana  two"    "Banana yellow "

Note that the resultant corpus are in a nested-list (reason why two sapply were used to view the content).

Loop through a tm corpus without losing corpus structure

1 Answers