Nothing
#' Sentence End Marks
#'
#'
#' \code{end_mark} - Grab the sentence end marks for a transcript. This can be
#' useful to categorize based on sentence type.
#'
#' @param text.var The text variable.
#' @param missing.end.mark A value to use for sentences with missing endmarks.
#' @param missing.text A value to use for sentences with missing (\code{NA})
#' text.
#' @param other.endmarks Other 1-2 character endmarks to search for.
#' @param grouping.var The grouping variables. Default \code{NULL} generates
#' one word list for all text. Also takes a single grouping variable or a list
#' of 1 or more grouping variables.
#' @param percent logical. If \code{TRUE} output given as percent. If
#' \code{FALSE} the output is proportion.
#' @param zero.replace Value to replace 0 values with.
#' @param digits Integer; number of decimal places to round when printing.
#' @param \ldots Other arguments passed to \code{end_mark}.
#' @rdname end_mark
#' @return Returns a character vector of qdap end marks for each sentence.
#' End marks include:
#' \item{"."}{Declarative sentence.}
#' \item{"?"}{Question sentence.}
#' \item{"!"}{Exclamatory sentence.}
#' \item{"|"}{Incomplete sentence.}
#' \item{"*."}{Imperative-declarative sentence.}
#' \item{"*?"}{Imperative-question sentence (unlikely to occur)}
#' \item{"*!"}{Imperative-exclamatory sentence.}
#' \item{"*|"}{Imperative-incomplete sentence.}
#' \item{"no.em"}{No end mark.}
#' \item{"blank"}{Empty cell/NA.}
#' @keywords end-mark
#' @export
#' @importFrom qdapTools matrix2df
#' @examples
#' \dontrun{
#' end_mark(DATA.SPLIT$state)
#' end_mark(mraja1spl$dialogue)
#' table(end_mark(mraja1spl$dialogue))
#' plot(end_mark(mraja1spl$dialogue))
#' ques <- mraja1spl[end_mark(mraja1spl$dialogue) == "?", ] #grab questions
#' htruncdf(ques)
#' non.ques <- mraja1spl[end_mark(mraja1spl$dialogue) != "?", ] #non questions
#' htruncdf(non.ques, 20)
#' ques.per <- mraja1spl[end_mark(mraja1spl$dialogue) %in% c(".", "?"), ] #grab ? and .
#' htruncdf(ques.per, 20)
#'
#' (x_by <- end_mark_by(DATA.SPLIT$state, DATA.SPLIT$person))
#' scores(x_by)
#' counts(x_by)
#' proportions(x_by)
#' preprocessed(x_by)
#' plot(scores(x_by))
#' plot(counts(x_by))
#' plot(proportions(x_by))
#' plot(preprocessed(x_by))
#'
#' #================================#
#' ## End Marks Over Time Examples ##
#' #================================#
#' ##EXAMPLE 1
#' sentpres <- lapply(with(pres_debates2012, split(dialogue, time)), function(x) {
#' end_mark(x)
#' })
#'
#' sentplots <- lapply(seq_along(sentpres), function(i) {
#' m <- plot(cumulative(sentpres[[i]]))
#' if (i != 2) m <- m + ylab("")
#' if (i != 3) m <- m + xlab(NULL)
#' m + ggtitle(paste("Debate", i))
#' })
#'
#' library(grid)
#' library(gridExtra)
#' do.call(grid.arrange, sentplots)
#'
#' ##EXAMPLE 2
#' sentraj <- lapply(with(rajSPLIT, split(dialogue, act)), function(x) {
#' end_mark(x)
#' })
#'
#' sentplots2 <- lapply(seq_along(sentraj), function(i) {
#' m <- plot(cumulative(sentraj[[i]]))
#' if (i != 2) m <- m + ylab("")
#' if (i != 3) m <- m + xlab(NULL)
#' act <- qcv(I, II, III, IV, V)
#' m + ggtitle(paste("Act", act[i]))
#' })
#'
#' ## ggplot2 function to extract legend
#' g_legend <- function(a.gplot){
#' tmp <- ggplot_gtable(ggplot_build(a.gplot))
#' leg <- which(sapply(tmp[["grobs"]], function(x) x[["name"]]) == "guide-box")
#' legend <- tmp[["grobs"]][[leg]]
#' legend
#' }
#'
#' ## remove legends from plots
#' sentplots3 <- lapply(sentplots2, function(x){
#' x + theme(legend.position="none") + xlab(NULL) + ylab(NULL)
#' })
#'
#' sentplots3[[6]] <- g_legend(sentplots2[[1]])
#'
#' do.call(grid.arrange, sentplots3)
#' }
end_mark <- function(text.var, missing.end.mark = "_", missing.text = NA,
other.endmarks = NULL) {
text.var <- as.character(text.var)
if (is.dp(text.var=text.var)){
warning(paste0("\n Some rows contain double punctuation.",
" Suggested use of sentSplit function."))
}
y <- nchar(text.var)
last1 <- substring(text.var, y)
last2 <- substring(text.var, y-1)
vals <- c("*.", "*?", "*!", "*|", other.endmarks)
for (i in seq_along(vals)) {
last1[last2 == vals[i]] <- vals[i]
}
last1[!last1 %in% c(vals, ".", "?", "!", "|")] <- missing.end.mark
last1[is.na(text.var)] <- missing.text
class(last1) <- c("end_mark", class(last1))
last1
}
#' Prints an end_mark object
#'
#' Prints an end_mark object
#'
#' @param x The end_mark object
#' @param \ldots ignored
#' @export
#' @method print end_mark
print.end_mark <-
function(x, ...) {
class(x) <- "character"
print(x)
}
#' Sentence End Marks
#'
#'
#' \code{end_mark_by} - Grab the sentence end marks for a transcript by grouping
#' variable(s).
#'
#' @rdname end_mark
#' @importFrom qdapTools mtabulate
#' @export
end_mark_by <- function(text.var, grouping.var, digits = 3, percent = FALSE,
zero.replace = 0, ...) {
if(is.null(grouping.var)) {
G <- "all"
} else {
if (is.list(grouping.var)) {
m <- unlist(as.character(substitute(grouping.var))[-1])
m <- sapply(strsplit(m, "$", fixed=TRUE), function(x) {
x[length(x)]
}
)
G <- paste(m, collapse="&")
} else {
G <- as.character(substitute(grouping.var))
G <- G[length(G)]
}
}
if(is.null(grouping.var)){
grouping <- rep("all", length(text.var))
} else {
if (is.list(grouping.var) & length(grouping.var)>1) {
grouping <- paste2(grouping.var)
} else {
grouping <- unlist(grouping.var)
}
}
DF <- data.frame(grouping, wc = wc(text.var), text.var,
check.names = FALSE, stringsAsFactors = FALSE)
DF[, "grouping"] <- factor(DF[, "grouping"])
DF[, "end.mark"] <- end_mark(DF[, "text.var"], ...)
DF2 <- mtabulate(split(DF[, "end.mark"], DF[, "grouping"]))
props <- prop(DF2, round=FALSE)
comb <- stats::setNames(data.frame(rownames(DF2),
raw_pro_comb(DF2, digits = digits, percent = percent,
zero.replace = zero.replace), check.names=FALSE),
c(G, colnames(props)))
out <- list(raw= stats::setNames(DF, c(G, colnames(DF[-1]))), count = matrix2df(DF2, G),
prop = matrix2df(props, G), rnp = comb)
class(out) <- c("end_mark_by", class(out))
attributes(out)[["digits"]] <- digits
attributes(out)[["percent"]] <- percent
attributes(out)[["zero.replace"]] <- zero.replace
out
}
#' Prints an end_mark_by object
#'
#' Prints an end_mark_by object
#'
#' @param x The end_mark_by object
#' @param \ldots ignored
#' @export
#' @method print end_mark_by
print.end_mark_by <-
function(x, ...) {
WD <- options()[["width"]]
options(width=3000)
print(x$rnp)
options(width=WD)
}
#' Question Counts
#'
#' View end_mark_by scores.
#'
#' end_mark_by Method for scores
#' @param x The \code{\link[qdap]{end_mark_by}} object.
#' @param \ldots ignored
#' @export
#' @method scores end_mark_by
scores.end_mark_by <- function(x, ...) {
out <- x[["rnp"]]
attributes(out) <- list(
class = c("end_mark_by_score", class(out)),
type = "end_mark_by_scores",
names = colnames(out),
row.names = rownames(out),
count = (x[["count"]])
)
out
}
#' Question Counts
#'
#' View end_mark_by counts.
#'
#' end_mark_by Method for counts
#' @param x The \code{\link[qdap]{end_mark_by}} object.
#' @param \ldots ignored
#' @export
#' @method counts end_mark_by
counts.end_mark_by <- function(x, ...) {
out <- x[["count"]]
attributes(out) <- list(
class = c("end_mark_by_count", class(out)),
type = "end_mark_by_counts",
names = colnames(out),
row.names = rownames(out)
)
out
}
#' Question Counts
#'
#' View \code{\link[qdap]{end_mark_by}} proportions.
#'
#' end_mark_by Method for proportions
#' @param x The end_mark_by object.
#' @param \ldots ignored
#' @export
#' @method proportions end_mark_by
proportions.end_mark_by <- function(x, ...) {
out <- x[["prop"]]
attributes(out) <- list(
class = c("end_mark_by_proportion", class(out)),
type = "end_mark_by_proportions",
names = colnames(out),
row.names = rownames(out)
)
out
}
#' Question Counts
#'
#' View \code{\link[qdap]{end_mark_by}} preprocessed.
#'
#' end_mark_by Method for preprocessed
#' @param x The end_mark_by object.
#' @param \ldots ignored
#' @export
#' @method preprocessed end_mark_by
preprocessed.end_mark_by <- function(x, ...) {
out <- x[["raw"]]
attributes(out) <- list(
class = c("end_mark_by_preprocessed", class(out)),
type = "end_mark_by_preprocessed",
names = colnames(out),
row.names = rownames(out)
)
out
}
#' Prints a end_mark_by_preprocessed object
#'
#' Prints a end_mark_by_preprocessed object
#'
#' @param x The end_mark_by_preprocessed object
#' @param \ldots ignored
#' @export
print.end_mark_by_preprocessed <-
function(x, ...) {
WD <- options()[["width"]]
options(width=3000)
class(x) <- "data.frame"
print(x)
options(width=WD)
}
#' Plots a end_mark_by_score Object
#'
#' Plots a end_mark_by_score object.
#'
#' @param x The end_mark_by_score object.
#' @param values logical. If \code{TRUE} the cell values will be included on
#' the heatmap.
#' @param \ldots Arguments passed to \code{\link[qdap]{qheat}}.
#' @method plot end_mark_by_score
#' @export
plot.end_mark_by_score <- function(x, values = TRUE, ...){
if (values) {
qheat(attributes(x)[["count"]], mat2 = x, ...)
} else {
qheat(attributes(x)[["count"]], ...)
}
}
#' Plots a end_mark_by_count Object
#'
#' Plots a end_mark_by_count object.
#'
#' @param x The end_mark_by_count object.
#' @param values logical. If \code{TRUE} the cell values will be included on
#' the heatmap.
#' @param \ldots Arguments passed to \code{\link[qdap]{qheat}}.
#' @method plot end_mark_by_count
#' @export
plot.end_mark_by_count <- function(x, values = TRUE, ...){
qheat(x, values = values, ...)
}
#' Plots a end_mark_by_proportion Object
#'
#' Plots a end_mark_by_proportion object.
#'
#' @param x The end_mark_by_proportion object.
#' @param values logical. If \code{TRUE} the cell values will be included on
#' the heatmap.
#' @param \ldots Arguments passed to \code{\link[qdap]{qheat}}.
#' @method plot end_mark_by_proportion
#' @export
plot.end_mark_by_proportion <- function(x, values = TRUE, ...){
qheat(x, values = values, ...)
}
#' Plots a end_mark_by_preprocessed Object
#'
#' Plots a end_mark_by_preprocessed object.
#'
#' @param x The end_mark_by_preprocessed object.
#' @param ncol The number of columns to use for \code{\link[ggplot2]{facet_wrap}}.
#' @param \ldots ignored
#' @importFrom ggplot2 xlab ggplot geom_bar facet_wrap coord_flip aes_string
#' @method plot end_mark_by_preprocessed
#' @export
plot.end_mark_by_preprocessed <- function(x, ncol = 1, ...){
group <- colnames(x)[1]
ord1 <- sort(table(x[, group] ))
x[, group] <- factor(x[, group], levels = names(ord1))
ord2 <- sort(table(x[, "end.mark"]), decreasing = TRUE)
x[, "end.mark"] <- factor(x[, "end.mark"], levels = names(ord2))
colnames(x)[1] <- "QDAP_GROUP"
ggplot(x, aes_string(x="QDAP_GROUP", fill="QDAP_GROUP")) +
geom_bar(show_guide = FALSE) +
coord_flip() +
facet_wrap(stats::reformulate("end.mark"), ncol = ncol) +
xlab(paste(unlist(sapply(unlist(strsplit(group, "&")),
Caps)), collapse =" & "))
}
#' Plots a end_mark_by Object
#'
#' Plots a end_mark_by object.
#'
#' @param x The end_mark_by object.
#' @param values logical. If \code{TRUE} the cell values will be included on
#' the heatmap.
#' @param \ldots Other arguments passed to \code{\link[qdap]{qheat}}.
#' @method plot end_mark_by
#' @export
plot.end_mark_by <- function(x, values = FALSE, ...) {
graphics::plot(scores(x), values = values, ...)
}
#' \code{cumulative.end_mark} - Generate end_mark over time (duration in
#' sentences).
#' @rdname cumulative
#' @export
#' @importFrom qdapTools %lc%
#' @method cumulative end_mark
cumulative.end_mark <- function(x, ...){
types_key <- structure(list(Symbol = c(".", "?", "!", "|", "*.", "*?", "*!",
"*|", "no.em", "blank"), Type = structure(c(1L, 10L, 2L, 7L,
3L, 6L, 4L, 5L, 9L, 8L), .Label = c("Declarative", "Exclamatory",
"Imperative-declarative", "Imperative-exclamatory", "Imperative-incomplete",
"Imperative-question", "Incomplete", "Missing", "No end mark",
"Question"), class = "factor")), .Names = c("Symbol", "Type"), row.names = c(NA,
-10L), class = "data.frame")
`%lcqdap%` <- qdapTools::`%lc%`
x <- x %lcqdap% types_key
x[is.na(x)] <- "Missing"
all <- names(table(x))
out <- lapply(1:length(x), function(i){
y <- c(table(x[1:i]))
missing <- all[!all %in% names(y)]
if(!identical(missing, character(0))){
y <- c(y, stats::setNames(rep(0, length(missing)), missing))
}
data.frame(t(y[as.character(types_key[["Type"]][types_key[["Type"]]
%in% names(y)])]), stringsAsFactors = FALSE)
})
plottingdf <- data.frame(Sentence = 1:length(out), do.call(rbind, out))
class(plottingdf) <- c("cumulative_end_mark", class(plottingdf))
plottingdf
}
#' Plots a cumulative_end_mark Object
#'
#' Plots a cumulative_end_mark object.
#'
#' @param x The cumulative_end_mark object.
#' @param \ldots ignored
#' @method plot cumulative_end_mark
#' @export
plot.cumulative_end_mark <- function(x, ...){
ord <- names(sort(colSums(x[nrow(x), -1, drop=FALSE]), decreasing=FALSE))
mplottingdf <- reshape2::melt(x, id="Sentence", variable="Type")
mplottingdf[, "Type"] <- factor(mplottingdf[, "Type"], levels=colnames(x)[-1])
ggplot2::ggplot(data=mplottingdf, ggplot2::aes_string(x="Sentence",
y="value", group="Type", color="Type")) + ggplot2::theme_bw() +
ggplot2::geom_line(size=1) +
ggplot2::ylab("Number") +
ggplot2::xlab("Duration") +
ggplot2::scale_x_continuous(expand = c(0, 0)) +
ggplot2::scale_colour_discrete(name = "Sentence Type")
}
#' Prints a cumulative_end_mark Object
#'
#' Prints a cumulative_end_mark object.
#'
#' @param x The cumulative_end_mark object.
#' @param \ldots ignored
#' @method print cumulative_end_mark
#' @export
print.cumulative_end_mark <- function(x, ...) {
print(plot.cumulative_end_mark(x, ...))
}
#' Plots an end_mark Object
#'
#' Plots an end_mark object.
#'
#' @param x The end_mark object.
#' @param \ldots ignored
#' @importFrom qdapTools %lc%
#' @method plot end_mark
#' @export
plot.end_mark <- function(x, ...) {
types_key <- structure(list(Symbol = c(".", "?", "!", "|", "*.", "*?", "*!",
"*|", "no.em", "blank"), Type = structure(c(1L, 10L, 2L, 7L,
3L, 6L, 4L, 5L, 9L, 8L), .Label = c("Declarative", "Exclamatory",
"Imperative-declarative", "Imperative-exclamatory", "Imperative-incomplete",
"Imperative-question", "Incomplete", "Missing", "No end mark",
"Question"), class = "factor")), .Names = c("Symbol", "Type"), row.names = c(NA,
-10L), class = "data.frame")
`%lcqdap%` <- qdapTools::`%lc%`
x <- x %lcqdap% types_key
x[is.na(x)] <- "Missing"
y <- table(x)
dat <- data.frame(y)
dat[, "x"] <- factor(dat[, "x"], levels=names(sort(y)))
ggplot2::ggplot(data=dat, aes_string(x="x", weight="Freq")) +
ggplot2::geom_bar() +
ggplot2::coord_flip() + ggplot2::ylab("Count") +
ggplot2::xlab("Sentence Type")
}
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.