Package Import

Load necessary packages and set one global option.

library(tidyverse)
library(pdftools)
library(tidytext)
library(knitr)
library(kableExtra)

Retrieve File

Download the file from the City of Raleigh website, read that file in as a character vector, and delete the downloaded file from the directory.

download.file("https://cityofraleigh0drupal.blob.core.usgovcloudapi.net/drupal-prod/COR11/FY2018AdoptedBudget20160612.pdf",
              "FY2018AdoptedBudget.pdf",
              mode = "wb")

txt = pdf_text("FY2018AdoptedBudget.pdf")

unlink("FY2018AdoptedBudget.pdf")

Create Data Frame

Create a page number character vector, create a data frame by binding the page number character vector with the extracted text, and finally “unnest” all of the page text into individual words.

page = as.character(1:length(txt))

df = data.frame(cbind(page, txt))

budget_words = df %>%
  mutate(txt = as.character(txt)) %>%
  unnest_tokens(word, txt)

Cleaning

Remove stop words and save as clean object, join sentiment lexicon with clean object, and group the object by page and sentiments before summarising.

cleaned = budget_words %>%
  anti_join(stop_words)

sentiment = cleaned %>%
  inner_join(get_sentiments("nrc"))

sent_count = sentiment %>%
  group_by(page, sentiment) %>%
  summarise(sent_count = n()) %>%
  ungroup() %>%
  mutate(page = as.integer(page))

Visualize

Negative Word Table

Trust Word Table