Using a list of word in findAssocs() in the tm R package

Question

I have been working with the findAssocs()function from the tm package in R. If I am using the function with a single word I don't have any problems and I can manually input a multiple words I would like to find associations to in the following format:

findAssoc(corpusname,"cat","dog","elephant",.75,.75,.75)

Again no problem with manually inputting the multiple terms. I am trying to find the associations to lists of terms sometimes that might contact 30 or 40 words. I would like to us either a list or vector with findAssocs() instead of having to type out each word every time. Any ideas how to do this? I tried making a custom function but I still so new to R I did not have any luck. Thanks.

Thanks for the help. R has a pretty steep learning curve for a newbie. I tried the first method that you suggested and get an the error "Error: is.character(terms) is not TRUE" The code that I am using is:

 #data for associates list
wordAssocList<- read.csv("Word Assocs List.txt")
# change TRUE to FALSE if you have no column headings in the CSV
as.character(wordAssocList)
attributes(wordAssocList)
my_assocs <- findAssocs(tdm, wordAssocList, .01)
my_assocs

For the output I get the following:

as.character(wordAssocList) [1] "logical(0)" attributes(wordAssocList) $names [1] "ÿþp"

$class [1] "data.frame"

$row.names integer(0)

my_assocs <- findAssocs(tdm, wordAssocList, .01) Error: is.character(terms) is not TRUE

phiver phiver · Accepted Answer · 2015-09-03T06:49:55

Vectors shouldn't be a problem. See following example.

library(tm)

data("crude")
tdm <- TermDocumentMatrix(crude)

words <- c("oil", "opec", "xyz")
corr <- c(0.7, 0.75, 0.1)

# returns a list
my_assocs <- findAssocs(tdm, words, corr)

# turns list into a list of named dataframes.
my_list <- lapply(my_assocs, function(x) data.frame(terms = names(x), cor = x, stringsAsFactors = FALSE))

edit: With the new version of dplyr (0.43) you can create a useful dataframe for the dataframes in the list, showing you the name of the dataframe the information is coming from. Handy for visualizations and other investigations.

my_df <- dplyr::bind_rows(my_list, .id = "source")

Source: local data frame [28 x 3]

   source    terms   cor
    (chr)    (chr) (dbl)
1     oil     15.8  0.87
2     oil  clearly  0.80
3     oil     late  0.80
4     oil   trying  0.80
5     oil      who  0.80
6     oil   winter  0.80
7     oil analysts  0.79
8     oil     said  0.78
9     oil  meeting  0.77
10    oil    above  0.76
..    ...      ...   ...

You could even use a dataframe instead of 2 vectors, just replace words and corr with the corresponding columns in your dataframe. The advantage of this, is that you can read in a text-file (or excel) where you have your lists of words and correlations

Using a list of word in findAssocs() in the tm R package

1 Answers