How to sample 75 percent of rows from a dtm?

How I can sample a dtm? I trying a lot of code but return me the same error

Error in dtm[splitter, ] : incorrect number of dimensions

This is the code:

n <- dtm$nrow
splitter <- sample(1:n, round(n * 0.75))
train_set <- dtm[splitter, ]
valid_set <- dtm[-splitter, ]

Solution

You can use the quanteda package for this. See example below:

Created data example based on the crude data set from tm:

library(tm)

data("crude")
crude <- as.VCorpus(crude)
crude <- tm_map(crude, stripWhitespace)
crude <- tm_map(crude, removePunctuation)
crude <- tm_map(crude, content_transformer(tolower))
crude <- tm_map(crude, removeWords, stopwords("english"))
crude <- tm_map(crude, stemDocument)

dtm <- DocumentTermMatrix(crude)


library(quanteda)

# Transform your dtm into a dfm for quanteda
my_dfm <- as.dfm(dtm)

# number of documents    
ndocs(my_dfm)
[1] 20

set.seed(4242)

# create training
train_set <- dfm_sample(my_dfm, 
                        size = round(ndoc(my_dfm) * 0.75),  # set sample size
                        margin = "documents")

# create test set by select the documents that do not match the documents in the training set.
test_set <- dfm_subset(my_dfm, !docnames(my_dfm) %in% docnames(train_set))

# number of documents in train
ndoc(train_set)
[1] 15

# number of documents in test
ndoc(test_set)
[1] 5

Afterwards you can use the quanteda function convert to convert your train and test sets to be used with topicmodels, lda, lsa, etc. See ?convert for more info.