I have a dtm and want to extract the top 5 terms by frequency for each document from the document term matrix.
I have a dtm built using the tm package
Terms
Docs aaaa aac abrt abused accept accepted
1 0 0 0 0 0 0
2 0 0 0 0 0 0
3 0 0 0 0 0 0
4 0 0 0 0 0 0
5 0 0 0 0 0 0
6 0 0 0 0 0 0
required output should be of the form:
Id
1 Term1 Term2 Term3 Term4 Term5
2 Term1 Term2 Term3 Term4 Term5
and so on for all the documents.
I have tried all the solutions available from stackoverflow ans other sources like Make dataframe of top N frequent terms for multiple corpora using tm package in R (converted to tdm and tried to bring to the output form but did not work)and others but noting seem to work.
Using Quanteda:
library(quanteda)
txt <- c("hello world world fizz", "foo bar bar buzz")
dfm <- dfm(txt)
topfeatures(dfm, n = 2, groups = seq_len(ndoc(dfm)))
# $`1`
# world hello
# 2 1
#
# $`2`
# bar foo
# 2 1
You can also convert between DocumentTermMatrix and dfm.
Or using the classical tm:
library(tm)
packageVersion("tm")
# [1] ‘0.7.1’
txt <- c(doc1="hello world world", doc2="foo bar bar fizz buzz")
dtm <- DocumentTermMatrix(Corpus(VectorSource(txt)))
n <- 5
(top <- findMostFreqTerms(dtm, n = n))
# $doc1
# world hello
# 2 1
#
# $doc2
# bar buzz fizz foo
# 2 1 1 1
do.call(rbind, lapply(top, function(x) { x <- names(x);length(x)<-n;x }))
# [,1] [,2] [,3] [,4] [,5]
# doc1 "world" "hello" NA NA NA
# doc2 "bar" "buzz" "fizz" "foo" NA
findMostFreqTerms is available since tm version 0.7-1.