CODEWITHSUNDEEP
rmatrixdocumentsimilarity
DI MI
DI MI

Reputation: 245

Compare Terms of Documents in Document Term Matrix in R

I need to build an Similarity Matrix by comparing terms of documents. So, for example, if Document1 and Document2 have 2 of the same terms, I need to write a 2 in my similarity matrix at m[1, 2]. My similarity matrix looks like this right now:

     [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9]
[1,]    0   NA   NA   NA   NA   NA   NA   NA   NA
[2,]    0    0   NA   NA   NA   NA   NA   NA   NA
[3,]    0    0    0   NA   NA   NA   NA   NA   NA
[4,]    0    0    0    0   NA   NA   NA   NA   NA
[5,]    0    0    0    0    0   NA   NA   NA   NA
[6,]    0    0    0    0    0    0   NA   NA   NA
[7,]    0    0    0    0    0    0    0   NA   NA
[8,]    0    0    0    0    0    0    0    0   NA

The documents and terms are inside a Document Term Matrix. Now I have to fill the similarity matrix by comparing all documents and their terms where it says NA in the similarity matrix. For every Term matched in an document pair I have to count +1 and inject the end value on the right place in the matrix.

My problem is, it seems I can't access the single documents and their terms inside the Document term Matrix. Is there any other way to perform this or am I missing something?

Here is the code:

install.packages("tm")
install.packages("openNLP")
install.packages("openNLPmodels.en")

Sys.setenv(NOAWT=TRUE)

library(tm)
library(openNLP)
library(openNLPmodels.en)

sample = c(
  "count eagle alien", 
  "dis bound eagle",   
  "bound count eagle dis",
  "count eagle dis alien",
  "bound eagle",
  "count dis alien",
  "bound count alien",
  "bound count",
  "count eagle dis"
)
print(sample)
corpus <- Corpus(VectorSource(sample))
inspect(corpus)

corpus <- tm_map(corpus, removeNumbers)
corpus <- tm_map(corpus, removePunctuation)
corpus <- tm_map(corpus, tolower)
corpus <- tm_map(corpus, removeWords, stopwords("english"))
corpus <- tm_map(corpus, stemDocument,language="english")
corpus <- tm_map(corpus, stripWhitespace)
corpus <- tm_map(corpus, tmTagPOS)
inspect(corpus)

dtm <- DocumentTermMatrix(corpus)
inspect(dtm)

# need to create similarity matrix here
#dist(dtm, method = "manhattan", diag = FALSE, upper = TRUE)

rowCount <- nrow(dtm)
similMatrix = matrix(nrow = rowCount - 1, ncol = rowCount)
show(similMatrix)
similMatrix[ row(similMatrix) >= col(similMatrix) ] <- 0

for(i in 1:(rowCount - 1)){  # rows
  for (j in i+1:rowCount){      # cols
  
      # need to compare document i and j here and write
      # the value into similarity matrix 
  }
}
show(similMatrix)

Upvotes: 1

Views: 4059

Answers (1)

user974514
user974514

Reputation: 552

I think you're missing one more row in your Similarity Matrix. Cause you don't get your last document covered. Mine is looking like this. 

     [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9]
 [1,]   NA   NA   NA   NA   NA   NA   NA   NA   NA
 [2,]    1   NA   NA   NA   NA   NA   NA   NA   NA
 [3,]    2    3   NA   NA   NA   NA   NA   NA   NA
 [4,]    3    2    3   NA   NA   NA   NA   NA   NA
 [5,]    1    2    2    1   NA   NA   NA   NA   NA
 [6,]    2    1    2    3    0   NA   NA   NA   NA
 [7,]    2    1    2    2    1    2   NA   NA   NA
 [8,]    1    1    2    1    1    1    2   NA   NA
 [9,]    2    2    3    3    1    2    1    1   NA

To get this result I did following steps.

mat=as.data.frame(as.matrix(dtm)) # you get the dataframe from DocumentTerm Matrix 
rowCount <- nrow(dtm)
colCount <- ncol(dtm)
similMatrix = matrix(nrow = rowCount, ncol = rowCount)
similMatrix[ row(similMatrix) >= col(similMatrix) ] <- 0
for(i in 1:(rowCount)){ #set all columns NA you can change to zeros if you need later
    similMatrix[i,i]=NA
} # then we will do the actual job
for(i in 1:rowCount ){  # rows
  for (j in 1:rowCount ){      # cols
      if(is.na(similMatrix[i,j])==F){
        a=mat[i,]
        b=mat[j,]
        for(k in 1:colCount){   #n number of Cols in Document term matrix

          if(a[k]==1 && a[k]==b[k]){
              similMatrix[i,j]=similMatrix[i,j]+1
          }
        }
      }
  }
}

Upvotes: 3

Related Questions