R tm 包：如何将文本与正参考词列表进行比较并返回正词出现次数

Question

What is the best approach to use the tm library to compare text to positive reference word list and return count of positive word occurrences I want to be able to return the sum of positive words in reference text.

Question: What is the best way to do this?

For example:

positiveword_list <- c("happy", "great", "fabulous", "great")

reference text:

exampleText <- c("ON A BRIGHT SPRING DAY in the year 1677, “the good ship 
Kent,” Captain Gregory Marlowe, Master, set sail from the great docks of London. She carried 230 English Quakers, outward bound for a new home in British North America. As the ship dropped down the Thames she was hailed by King Charles II, who happened to be sailing on the river. The two vessels made a striking contrast. The King’s yacht was sleek and proud in gleaming paintwork, with small cannons peeping through wreaths of gold leaf, a wooden unicorn prancing high above her prow, and the royal arms emblazoned upon her stern. She seemed to dance upon the water— new sails shining white in the sun, flags streaming bravely from her mastheads, officers in brilliant uniform, ladies in court costume, servants in livery, musicians playing, and spaniels yapping. At the center of attention was the saturnine figure of the King himself in all his regal splendor. On the other side of the river came the emigrant ship. She would have been bluff-bowed and round-sided, with dirty sails and a salt-stained hull, and a single ensign drooping from its halyard. Her bulwarks were lined with apprehensive passengers— some dressed in the rough gray homespun of the northern Pen-nines, others in the brown drab of London tradesmen, several in the blue suits of servant-apprentices, and a few in the tattered motley of the country poor.")

Here is some background:

What I am trying to do is count the number of positive works and store the count in a dataframe as a new column.

count <-    length(which(lapply(positiveword_list, grepl, x = exampleText]) == TRUE))

thus:

dataframeIn %>% mutate( posCount <- (length(which(lapply(positiveword_list, grepl, x = text) == TRUE)))) 

where text is a column in dataFrameIn (ie dataFrameIn$text)

Answer 1

You can do this without using tm package.

Try this

contained <- lapply(positiveword_list, grepl, x = exampleText)

lapply returns a list.

Words present:

>positiveword_list[contained == T]
"great" "great"
>length(contained[contained==T])
2

Words not present:

>positiveword_list[contained == F]
"happy"    "fabulous"
>length(contained[contained==F])
2

Answer 2

Here's another method using a custom-built tool, where you can define a dictionary of positive words and apply this to any number of texts, to count the positive key words. This uses the quanteda package and the dfm() method to create a document-feature matrix, with the dictionary = argument. (See ?dictionary .)

require(quanteda)
posDic <- dictionary(list(positive = positiveword_list))
myDfm <- dfm(exampleText, dictionary = posDic)
# Creating a dfm from a character vector ...
# ... lowercasing
# ... tokenizing
# ... indexing documents: 1 document
# ... indexing features: 157 feature types
# ... applying a dictionary consisting of 1 key
# ... created a 1 x 1 sparse dfm
# ... complete. 
# Elapsed time: 0.014 seconds.

as.data.frame(myDfm)
#       positive
# text1        1

# produces a data frame with the text and the positive count
cbind(text = exampleText, as.data.frame(myDfm))

Note: This is probably not important to the example, but the usage of "great" in the exampleText is not as a positive word. Illustrates the perils of polysemy and dictionaries.