hide
# Convert text into a data frame
little_women_tidy <- data.frame(little_women_text) %>%
mutate(text_full = str_split(little_women_text, pattern = '\\n')) %>% # break pages up into individual lines
unnest(text_full) %>% # convert each line of text into one row
mutate(text_full = str_trim(text_full)) %>% # remove excess whitespace
mutate(text_full = str_squish(text_full))
# Organize strings by chapter number
little_women_df <- little_women_tidy %>%
slice(-(1:101)) %>% # remove first rows containing preface and table of contents, start at Ch. 1
mutate(chapter = case_when(
str_detect(text_full, "CHAPTER") ~ text_full, # detect rows containing the word "Chapter"
TRUE ~ NA_character_
)) %>%
fill(chapter) %>% # fill in Chapter # for each row in respective chapter
separate(col = chapter, into = c("ch", "no"), sep = " ") %>% # separate Chapter and # into two columns
mutate(no = str_remove(no, pattern = "\\.")) %>% # remove period after each Chapter#
mutate(no = str_replace(no, pattern = "in", replacement = "III")) %>% # fix error (XVIII was being read in as XVin)
mutate(chapter = as.numeric(as.roman(no))) # convert Ch# to be recognized as roman numerals and numeric class
hide
# Tokenization - wrangling words/tokens into tidy format
little_women_tokens <- little_women_df %>%
unnest_tokens(word, text_full) %>% # split a column into tokens (words) with one token per row
dplyr::select(-little_women_text) %>% # remove little women text column
anti_join(stop_words) %>% # remove stop words (e.g. a, of, the, etc.)
mutate(word = str_squish(word)) %>%
filter(word != "jo") %>% # remove character names
filter(word != "jo's") %>%
filter(word != "beth") %>%
filter(word != "meg") %>%
filter(word != "amy") %>%
filter(word != "amy's") %>%
filter(word != "march") %>%
filter(word != "laurie") %>%
filter(word != "john") %>%
filter(word != "brooke") %>%
filter(word != "hannah") %>%
filter(word != "kate") %>%
filter(word != "demi") %>%
filter(word != "bhaer") %>%
filter(word != "fred") %>%
filter(word != "women") %>%
filter(word != "chapter") %>%
filter(word != "do") %>% # remove misc. words
filter(word != "n't") %>%
filter(word != "11") %>%
filter(word != "ve") %>%
filter(word != "ing")
Most Frequently Used Words in Little Women
hide
# Get word counts
little_women_wordcount <- little_women_tokens %>%
count(word)
# Make a word cloud using the top 100 words
top100 <- little_women_wordcount %>%
arrange(-n) %>% # arrange counts from largest to smallest
slice(1:100) # keep top 100 words
word_cloud <- ggplot(data = top100, aes(label = word)) +
geom_text_wordcloud(aes(color = n, size = n), shape = "diamond") +
scale_size_area(max_size = 6) +
scale_color_gradientn(colors = c("blue","blueviolet","violetred")) +
theme_minimal()
word_cloud
Figure 1. Wordcloud containing the top 100 most frequently used words in Louisa May Alcott’s Little Women. Character names and stop words, such as “is”, “of”, “the”, and “and”, were removed from word counts to allow for more interesting word assortment. (Data: Alcott 1916).
Sentiment Analysis of Little Women
hide
# Get word counts by chapter
little_women_wordcount_ch <- little_women_tokens %>%
count(chapter, word)
# Bind words in `little_women_wordcount` to the `afinn` lexicon
little_women_afinn <- little_women_wordcount_ch %>%
inner_join(get_sentiments("afinn"))
# Find the mean AFINN score by chapter:
afinn_means <- little_women_afinn %>%
group_by(chapter) %>%
summarize(mean_afinn = mean(value)) %>%
mutate(pos = mean_afinn >= 0) # create a new column that indicates + or - value
# Graph positivity and negativity of chapters
ggplot(data = afinn_means,
aes(x = fct_rev(as.factor(chapter)),
y = mean_afinn,
fill = pos)) +
geom_col(position = "identity") +
scale_fill_manual(values = c("indianred", "cornflowerblue")) +
scale_x_discrete(breaks = c(5, 10, 15, 20, 25, 30, 35, 40, 45)) +
scale_y_continuous(limits = c(-0.5, 1.2)) +
coord_flip() +
theme_light() +
labs(x = "Chapter",
y = "Mean AFINN sentiment score",
title = "Little Women - Sentiment Analysis by Chapter") +
theme(legend.position = "none")
Figure 2. Column graph displaying mean AFINN scores for the words in Parts 1 and 2 (Chapters 1-47) of Little Women, using the AFINN lexicon. The AFINN lexicon ranks words from -5 (very negative sentiment) to +5 (very positive sentiment). Blue bars indicate chapters containing more words (on average) with positive mean sentiments, while red bars indicate chapters containing more words (on average) with negative mean sentiments. (Data: Alcott 1916, Nielson 2011).
Citations:
Text Data: Alcott, Louisa M. Little Women: or Meg, Jo, Beth, and Amy. Little, Brown, and Company, 1916.