This code snippet takes a vector of strings and calculates the percentage of passive voice in the input text. It uses Stanford NLP tool and coreNLP for R.

passive.R

library(rJava)
library(coreNLP)
initCoreNLP()

#in this case, 'test' is a data frame with a col named 'text'

for (i in 1:dim(test)[1]) {
  cat(paste0(i / dim(test)[1] * 100, '% completed'))
  ann <- annotateString(paste(test$text[i]), format = c("obj"), outputFile = NA, includeXSL = FALSE)
  gd <- getDependency(ann)
  passive_sub <- length(gd$type[gd$type=="nsubjpass"])/length(gd$type[gd$type!="punct"])
  passive_aux <- length(gd$type[gd$type=="auxpass"])/length(gd$type[gd$type!="punct"])
  passive_age <- length(gd$type[gd$type=="nmod:agent"])/length(gd$type[gd$type!="punct"])
  
  # here i create three metrics of passives in the input data frame. i prefer using passive_aux as my main metric
  test$passive_sub[i] <- as.numeric(passive_sub)
  test$passive_aux[i] <- as.numeric(passive_aux)
  test$passive_age[i] <- as.numeric(passive_age) #passives with by-phrases

  if (i == dim(test)[1]) cat(': Done')
  else cat('\014')
}