Quanteda - Extracting identified dictionary words

问题:

I am trying to extract the identified dictionary words from a Quanteda dfm, but have been unable to find a solution.

Does someone have a solution for this?

Sample input:

dict <- dictionary(list(season = c("spring", "summer", "fall", "winter")))
dfm  <- dfm("summer is great", dictionary  = dict)

Output:

 > dfm
 Document-feature matrix of: 1 document, 1 feature.
 1 x 1 sparse Matrix of class "dfmSparse"

   features
docs    season
text1      1

I now know that a seasonality dict word has been identified in the sentence, but I would also like to know which word it was.

This should preferably be extracted in the table format:

docs    dict     dictWord
text1   season   summer

回答1:

You can create a second dfm using the keptFeatures argument, and then cbind() it to the first, dictionaried dfm.

dict <- dictionary(list(season = c("spring", "summer", "fall", "winter")))
txt <- "summer is great"
season_dfm  <- dfm(txt, dictionary  = dict, verbose = FALSE)
dict_dfm <- dfm(txt, select = dict, verbose = FALSE)

cbind(season_dfm, dict_dfm)
## Document-feature matrix of: 1 document, 2 features.
## 1 x 2 sparse Matrix of class "dfmSparse"
##       season summer
## text1      1      1

If you want the output as a table, it would be:

dict_df <- as.data.frame(combined_dfm)
names(dict_df)[2] <- "dictWord"
dict_df
##       season dictWord
## text1      1        1

but that only works if you have a single dictionary value per text -- otherwise the dict_dfm will have multiple feature columns.