Quanteda package provides the sparse document-feature matrix DFM and its methods contain removeFeatures. I have tried dfm(x, removeFeatures="\\b[a-z]{1-3}\\b") to remove too short words as well as dfm(x, keptFeatures="\\b[a-z]{4-99}\\b") to preserve sufficiently long words but not working, basically doing the same thing i.e. removing too short words.
How can I remove a regex match from a Quanteda DFM object?
Example.
myMatrix <-dfm(myData, ignoredFeatures = stopwords("english"),
stem = TRUE, toLower = TRUE, removeNumbers = TRUE,
removePunct = TRUE, removeSeparators = TRUE, language = "english")
#
#How to use keptFeatures/removeFeatures here?
#Instead of RemoveFeatures/keptFeatures methods, I tried it like this but not working
x<-unique(gsub("\\b[a-zA-Z0-9]{1,3}\\b", "", colnames(myMatrix)));
x<-x[x!=""];
mmyMatrix<-myMatrix;
colnames(mmyMatrix) <- x
Sample DFM
myData <- c("a aothu oat hoah huh huh huhhh h h h n", "hello h a b c d abc abcde", "hello hallo hei hej", "Hello my name is hhh.")
myMatrix <- dfm(myData)
It's dfm_select, in >= v0.9.9:
myMatrix
## Document-feature matrix of: 4 documents, 22 features (70.5% sparse).
dfm_select(myMatrix, "\\b[a-zA-Z0-9]{1,3}\\b", selection = "keep", valuetype = "regex")
## kept 14 features, from 1 supplied (regex) feature types
## Document-feature matrix of: 4 documents, 14 features (71.4% sparse).
## 4 x 14 sparse Matrix of class "dfmSparse"
## features
## docs a oat huh h n b c d abc hei hej my is hhh
## text1 1 1 2 3 1 0 0 0 0 0 0 0 0 0
## text2 1 0 0 1 0 1 1 1 1 0 0 0 0 0
## text3 0 0 0 0 0 0 0 0 0 1 1 0 0 0
## text4 0 0 0 0 0 0 0 0 0 0 0 1 1 1