class_update; dfm_gen; merger: updated functions to accept text parameter for both old style 'lemmas' and new style 'ud'

R/class_update.R

@@ -6,6 +6,7 @@
 #' @param model_final The classification model (e.g. output from textstat_nb(), svm() or others)
 #' @param dfm_words A dfm containing all the words and only the words used to generate the model (is used for subsetting)
 #' @param varname String containing the variable name to use for the classification result, usually has the format computerCodes.varname
+#' @param text String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"
 #' @param es_super Password for write access to ElasticSearch
 #' @return As this is a nested function used within elasticizer, there is no return output
 #' @export
@@ -14,9 +15,9 @@
 #################################################################################################
 #################################### Update any kind of classification ##########################
 #################################################################################################
-class_update <- function(out, localhost = T, model_final, dfm_words, varname, es_super = .rs.askForPassword('ElasticSearch WRITE')) {
+class_update <- function(out, localhost = T, model_final, dfm_words, varname, text, es_super = .rs.askForPassword('ElasticSearch WRITE')) {
   print('updating')
-  dfm <- dfm_gen(out, text = 'lemmas') %>%
+  dfm <- dfm_gen(out, text = text) %>%
     dfm_keep(dfm_words, valuetype="fixed", verbose=T)
   pred <- data.frame(id = out$`_id`, pred = predict(model_final, newdata = dfm))
   bulk <- apply(pred, 1, bulk_writer, varname = varname, type = 'set')

R/dfm_gen.R

@@ -3,7 +3,7 @@
 #' Generates dfm from ElasticSearch output
 #' @param out The elasticizer-generated data frame
 #' @param words String indicating the number of words to keep from each document (maximum document length), 999 indicates the whole document
-#' @param text String indicating whether the "merged" field will contain the "full" text, or "lemmas"
+#' @param text String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"
 #' @return A Quanteda dfm
 #' @export
 #' @examples
@@ -21,8 +21,8 @@ dfm_gen <- function(out, words = '999', text = "lemmas") {
   out <- out %>%
     select(`_id`, matches("_source.*")) ### Keep only the id and anything belonging to the source field
   fields <- length(names(out))
-  if (text == "lemmas") {
-    out$merged <- unlist(mclapply(seq(1,length(out[[1]]),1),merger, out = out, mc.cores = detectCores()))
+  if (text == "lemmas" || text == 'ud') {
+    out$merged <- unlist(mclapply(seq(1,length(out[[1]]),1),merger, out = out, text = text, mc.cores = detectCores()))
   }
   if (text == "full") {
     out$merged <- str_c(str_replace_na(out$`_source.title`, replacement = " "),

R/merger.R

@@ -4,18 +4,24 @@
 #' @param row A row number form the Elasticizer-generated data frame
 #' @param words String indicating the number of words to keep from each document (maximum document length), 999 indicates the whole document
 #' @param out The elasticizer-generated data frame
+#' @param text String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"
 #' @return A documentified string of lemmas, one document at a time
 #' @export
 #' @examples
-#' merger(1, words = '999', out = out)
+#' merger(1, words = '999', out, text)
 #################################################################################################
 #################################### Reconstructing documents from lemmas########################
 #################################################################################################
 ## Only merging lemmas for now, feature selection has no impact on junk classification
-merger <- function(row, out = out) {
+merger <- function(row, out, text) {
   df <- out[row,]
   # Mergin lemmas into single string
-  lemmas <- paste(str_split(df$`_source.tokens.lemmas`, "\\|")[[1]],collapse = ' ')
+  if (text == 'lemmas') {
+    lemmas <- paste(str_split(df$`_source.tokens.lemmas`, "\\|")[[1]],collapse = ' ')
+  }
+  if (text == 'ud') {
+    lemmas <- paste0(df$`_source.ud`[[1]]$lemma[[1]], collapse = ' ')
+  }
   # Replacing $-marked punctuation with their regular forms
   lemmas <- str_replace_all(lemmas," \\$(.+?)", "\\1") %>%
     ### Removing numbers and non-words containing numbers

man/class_update.Rd

@@ -4,7 +4,7 @@
 \alias{class_update}
 \title{Classifier function for use in combination with the elasticizer function as 'update' parameter (without brackets), see elasticizer documentation for more information}
 \usage{
-class_update(out, localhost = T, model_final, dfm_words, varname,
+class_update(out, localhost = T, model_final, dfm_words, varname, text,
   es_super = .rs.askForPassword("ElasticSearch WRITE"))
 }
 \arguments{
@@ -18,6 +18,8 @@ class_update(out, localhost = T, model_final, dfm_words, varname,
 
 \item{varname}{String containing the variable name to use for the classification result, usually has the format computerCodes.varname}
 
+\item{text}{String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"}
+
 \item{es_super}{Password for write access to ElasticSearch}
 }
 \value{

man/dfm_gen.Rd

@@ -11,7 +11,7 @@ dfm_gen(out, words = "999", text = "lemmas")
 
 \item{words}{String indicating the number of words to keep from each document (maximum document length), 999 indicates the whole document}
 
-\item{text}{String indicating whether the "merged" field will contain the "full" text, or "lemmas"}
+\item{text}{String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"}
 }
 \value{
 A Quanteda dfm

man/merger.Rd

@@ -4,13 +4,15 @@
 \alias{merger}
 \title{Merges list of lemmas back into a pseudo-document}
 \usage{
-merger(row, out = out)
+merger(row, out, text)
 }
 \arguments{
 \item{row}{A row number form the Elasticizer-generated data frame}
 
 \item{out}{The elasticizer-generated data frame}
 
+\item{text}{String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"}
+
 \item{words}{String indicating the number of words to keep from each document (maximum document length), 999 indicates the whole document}
 }
 \value{
@@ -20,5 +22,5 @@ A documentified string of lemmas, one document at a time
 Merges list of lemmas back into a pseudo-document
 }
 \examples{
-merger(1, words = '999', out = out)
+merger(1, words = '999', out, text)
 }