Sift through each row in a dataframe and manually classify it - r

Question

Can someone recommend an efficient way to sift through each row in a dataframe and manually classify it? For example I might be wanting to separate spam from e-mail, or shortlist job adverts, job applicants, or dating agency profiles (I understand Tinder does this by getting you to swipe left or right).
My dataset is small enough to classify manually. I suppose if it was larger I might only want to manually classify a portion of it in order to train a machine-learning algorithm such as Naive Bayes to finish the task for me.
I'll show you what I've got at the moment, but this isn't a particularly original task, so there must be a less crude way to do this that someone has already thought of! (As a newcomer, I'm impressed by the power of R, but also nonplussed when little tasks like clearing the screen or capturing a keystroke turn out to be non-trivial)
# Let us suppose I am using this built-in dataset to draw up a
# shortlist of where I might wish to go on holiday
df <- data.frame(state.x77);
# pp - define a task-specific pretty print function
pp <- function(row) {
print(row); # Example dataset is simple enough to just print the entire row
}
# cls - clear the screen (this hack works on Windows but I've commented it for now)
cls <- function() {
#system("powershell -ExecutionPolicy Bypass -command (New-Object -ComObject Wscript.Shell).SendKeys([string][char]12)");
}
# It would halve the number of keystrokes needed if I knew a way to read
# a single character
readcharacter <- readline;
sift <- function(df, pp)
{
classification = rep('', nrow(df));
for (nRow in 1:nrow(df))
{
cls();
pp(df[nRow,]);
cat("\nEnter 'a' to discard, 'd' to keep, 'q' to quit\n");
char <- '';
while (char != 'a' && char != 'd' && char != 'q') {
char <- readcharacter();
}
if (char == 'q')
break;
classification[nRow] = char;
}
return(cbind(df,classification=classification));
}
result = sift(df, pp);
cls();
cat("Shortlist:\n");
print(row.names(result[result$classification=='d',]));

Answer 1

So how does the StackOverflow community feel about me using this Shiny app to solve my problem? I wouldn't expect to see Shiny used in this early part of data analysis - normally it only comes into play once we have some results we'd like to explore or present dynamically.
Learning Shiny was fun and useful, but I'd much prefer it if a less complicated answer could be found.
library(shiny);
#
# shortlist - function that allows us to shortlist through the rows in a data frame efficiently
#
shortlist <- function(df, sTitle, sRowName) {
createUI <- function() {
listHeading <- list(
textOutput(outputId = "Progress"),
tags$br(),
fluidRow(
column(width=1, sRowName),
column(width=9, textOutput(outputId = "RowName"))));
listFields <- lapply(names(df), function(sFieldname) {
return(fluidRow(
column(width=1, sFieldname),
column(width=9, textOutput(outputId = sFieldname))));
});
listInputs <- list(
tags$br(),
tags$table(
tags$tr(
tags$td(" "),
tags$td(actionButton(inputId="Up", label="W", disabled=TRUE, width="100%"))),
tags$tr(
tags$td(width="100px", actionButton(inputId="Discard", label="Discard, A", width="100%")),
tags$td(width="100px", actionButton(inputId="Down", label="S", disabled=TRUE, width="100%")),
tags$td(width="100px", actionButton(inputId="Keep", label="Keep, D", width="100%")))),
tags$script("
// JavaScript implemented keyboard shortcuts, including lots of conditions to
// ensure we're finished processing one keystroke before we start the next.
var bReady = false;
$(document).on('shiny:recalculating', function(event) {
bReady = false;
});
$(document).on('shiny:recalculated', function(event) {
setTimeout(function() {bReady = true;}, 500);
});
$(document).on('keypress', function(event) {
if (bReady) {
switch(event.key.toLowerCase()) {
case 'a':
document.getElementById('Discard').click();
bReady = false;
break;
case 'd':
document.getElementById('Keep').click();
bReady = false;
break;
}
}
});
// End of JavaScript
"));
listPanel <- list(
title = sTitle,
tags$br(),
conditionalPanel(
condition = paste("input.Keep + input.Discard <", nrow(df)),
append(append(listHeading, listFields), listInputs)));
listShortlist <- list(
tags$hr(),
tags$h4("Shortlist:"),
dataTableOutput(outputId="Shortlist"));
ui <- do.call(fluidPage, append(listPanel, listShortlist));
return(ui);
}
app <- shinyApp(ui = createUI(), server = function(input, output) {
classification <- rep('', nrow(df));
getRow <- reactive({
return (input$Keep + input$Discard + 1);
});
classifyRow <- function(nRow, char) {
if (nRow <= nrow(df)) {
classification[nRow] <<- char;
}
# In interactive mode, automatically stop the app when we're finished
if ( interactive() && nRow >= nrow(df) ) {
stopApp(classification);
}
}
observeEvent(input$Discard, {classifyRow(getRow() - 1, 'a')});
observeEvent(input$Keep, {classifyRow(getRow() - 1, 'd')});
output$Progress = renderText({paste("Showing record", getRow(), "of", nrow(df))});
output$RowName = renderText({row.names(df)[getRow()]});
lapply(names(df), function(sFieldname) {
output[[sFieldname]] <- renderText({df[getRow(), sFieldname]});
});
output$Shortlist <- renderDataTable(options = list(paging = FALSE, searching = FALSE), {
# Mention the 'keep' input to ensure this code is called when the 'keep' button
# is pressed. That way the shortlist gets updated when an item to be added to it.
dummy <- input$Keep;
# Construct the shortlist
shortlist <- data.frame(row.names(df[classification == 'd',]));
colnames(shortlist) <- sRowName;
return(shortlist);
});
});
if (interactive()) {
classification <- runApp(app);
return(cbind(df, classification = classification));
} else {
return(app);
}
}
#
# And now some example code.
# Shortlist the built in state.x77 data set (let us suppose I am drawing up
# a shortlist of where I might wish to go on holiday)
#
df <- data.frame(state.x77);
result <- shortlist(df = df, "Choose states", "State");
if (interactive()) {
cat("Shortlist:\n");
print(row.names(result[result$classification == 'd',]));
} else {
return (result);
}