R stemming a string/document/corpus

Question

I'm trying to do some stemming in R but it only seems to work on individual documents. My end goal is a term document matrix that shows the frequency of each term in the document.

Here's an example:

require(RWeka)
require(tm)
require(Snowball)

worder1<- c("I am taking","these are the samples",
"He speaks differently","This is distilled","It was placed")
df1 <- data.frame(id=1:5, words=worder1)

> df1
  id                 words
1  1           I am taking
2  2 these are the samples
3  3 He speaks differently
4  4     This is distilled
5  5         It was placed

This method works for the stemming part but not the term document matrix part:

> corp1 <- Corpus(VectorSource(df1$words))
> inspect(corp1)
A corpus with 5 text documents

The metadata consists of 2 tag-value pairs and a data frame
Available tags are:
  create_date creator 
Available variables in the data frame are:
  MetaID 

[[1]]
I am taking

[[2]]
these are the samples

[[3]]
He speaks differently

[[4]]
This is distilled

[[5]]
It was placed

> corp1 <- tm_map(corp1, SnowballStemmer)
> inspect(corp1)
A corpus with 5 text documents

The metadata consists of 2 tag-value pairs and a data frame
Available tags are:
  create_date creator 
Available variables in the data frame are:
  MetaID 

[[1]]
[1] I am tak

[[2]]
[1] these are the sampl

[[3]]
[1] He speaks differ

[[4]]
[1] This is distil

[[5]]
[1] It was plac

>  class(corp1)
[1] "VCorpus" "Corpus"  "list"   
> tdm1 <- TermDocumentMatrix(corp1)
Error in UseMethod("Content", x) : 
  no applicable method for 'Content' applied to an object of class "character"

So instead I tried creating the term document matrix first but this time the words don't get stemmed:

> corp1 <- Corpus(VectorSource(df1$words))
> tdm1 <- TermDocumentMatrix(corp1, control=list(stemDocument=TRUE))
>  as.matrix(tdm1)
             Docs
Terms         1 2 3 4 5
  are         0 1 0 0 0
  differently 0 0 1 0 0
  distilled   0 0 0 1 0
  placed      0 0 0 0 1
  samples     0 1 0 0 0
  speaks      0 0 1 0 0
  taking      1 0 0 0 0
  the         0 1 0 0 0
  these       0 1 0 0 0
  this        0 0 0 1 0
  was         0 0 0 0 1

Here the words are obviously not stemmed.

Any suggestions?

Solution 1

The RTextTools package on CRAN allows you to do this.

library(RTextTools)
worder1<- c("I am taking","these are the samples",
"He speaks differently","This is distilled","It was placed")
df1 <- data.frame(id=1:5, words=worder1)

matrix <- create_matrix(df1, stemWords=TRUE, removeStopwords=FALSE, minWordLength=2)
colnames(matrix) # SEE THE STEMMED TERMS

This returns a DocumentTermMatrix that can be used with package tm. You can play around with the other parameters (e.g. removing stopwords, changing the minimum word length, using a stemmer for a different language) to get the results you need. When displayed as.matrix the example produces the following term matrix:

                         Terms
Docs                      am are differ distil he is it place sampl speak take the these this was
  1 I am taking            1   0      0      0  0  0  0     0     0     0    1   0     0    0   0
  2 these are the samples  0   1      0      0  0  0  0     0     1     0    0   1     1    0   0
  3 He speaks differently  0   0      1      0  1  0  0     0     0     1    0   0     0    0   0
  4 This is distilled      0   0      0      1  0  1  0     0     0     0    0   0     0    1   0
  5 It was placed          0   0      0      0  0  0  1     1     0     0    0   0     0    0   1

Solution 2

This works in R as expected with tm version 0.6. You had a few minor errors that prevented the stemming for working correctly, perhaps they're from an older version of tm? Anyway, here's how to make it work:

require(RWeka)
require(tm)

The stemming package is not your Snowball but SnowballC:

require(SnowballC)

worder1<- c("I am taking","these are the samples",
            "He speaks differently","This is distilled","It was placed")
df1 <- data.frame(id=1:5, words=worder1)
corp1 <- Corpus(VectorSource(df1$words))
inspect(corp1)

Change your SnowballStemmer to stemDocument in the next line like so:

corp1 <- tm_map(corp1, stemDocument)
inspect(corp1)

Words are stemmed, as expected:

<<VCorpus (documents: 5, metadata (corpus/indexed): 0/0)>>

[[1]]
<<PlainTextDocument (metadata: 7)>>
I am take

[[2]]
<<PlainTextDocument (metadata: 7)>>
these are the sampl

[[3]]
<<PlainTextDocument (metadata: 7)>>
He speak differ

[[4]]
<<PlainTextDocument (metadata: 7)>>
This is distil

[[5]]
<<PlainTextDocument (metadata: 7)>>
It was place

Now do the term document matrix:

corp1 <- Corpus(VectorSource(df1$words))

Change your stemDocument to stemming:

tdm1 <- TermDocumentMatrix(corp1, control=list(stemming=TRUE))
as.matrix(tdm1)

And we get a tdm of stemmed words, as expected:

        Docs
Terms    1 2 3 4 5
  are    0 1 0 0 0
  differ 0 0 1 0 0
  distil 0 0 0 1 0
  place  0 0 0 0 1
  sampl  0 1 0 0 0
  speak  0 0 1 0 0
  take   1 0 0 0 0
  the    0 1 0 0 0
  these  0 1 0 0 0
  this   0 0 0 1 0
  was    0 0 0 0 1

So there you go. Perhaps a more careful reading of the tm docs might have saved a bit of your time with this ;)

Solution 3

Yes for steming words of document in a Corpus you required Rweka, Snowball and tm package.

use following instruction

> library (tm)
#set your directory Suppose u have set "F:/St" then next command is 
> a<-Corpus(DirSource("/st"), 
            readerControl=list(language="english")) # "/st" it is path of your directory
> a<-tm_map(a, stemDocument, language="english")
> inspect(a)

sure you will find your desired result.