R/word_corr.R

In evan-l-munson/SAoTD: Sentiment Analysis of Twitter Data

#' @title Twitter Word Correlations
#'
#' @description The word correlation displays the mutual relationship between 
#'   words.
#'
#' @param DataFrameTidy Data Frame of Twitter Data that has been tidy'd.
#' @param number The number of word instances to be included.
#' @param sort Rank order the results from most to least correlated.
#'
#' @importFrom dplyr group_by filter quo n
#' @importFrom widyr pairwise_cor
#'
#' @return A Tibble.
#'
#' @examples 
#' \dontrun{
#' library(saotd)
#' data <- raw_tweets
#' tidy_data <- Tidy(DataFrame = data)
#' TD_Word_Corr <- word_corr(DataFrameTidy = tidy_data,
#'                           number = 500,
#'                           sort = TRUE)
#'
#' TD_Word_Corr
#' }                    
#' @export

word_corr <- function(DataFrameTidy, 
                      number, 
                      sort = TRUE) {

  # input checks
  if (!is.data.frame(DataFrameTidy)) {
    stop("The input for this function is a data frame.")
  }

  if (!(("Token" %in% colnames(DataFrameTidy)) & 
       ("key" %in% colnames(DataFrameTidy)))) {
    stop("The data frame is not properly constructed.  
         The data frame must contain at minimum the columns: Token and key.")
  }

  if (number <= 1) {
    stop("Must choose number of Correlation pairs greater than 1.")
  }

  # configure defusing operators for packages checking
  Token <- dplyr::quo(Token)
  n <- dplyr::quo(n)
  key <- dplyr::quo(key)

  # function main body
  # suppress Warning message:
  #   `tbl_df()` was deprecated in dplyr 1.0.0.
  #   Please use `tibble::as_tibble()` instead.
  #   Currently `widyr` v0.1.3 has not accounted for the `dplyr` v1.0.0 update.
  TD_Word_Correlation <- DataFrameTidy %>%
    dplyr::group_by(Token) %>%
    dplyr::filter(dplyr::n() >= number) %>%
    widyr::pairwise_cor(Token, key, sort = sort)

  return(TD_Word_Correlation)

}