Install the package:
install.packages("Goodreader")
And load the package:
library(Goodreader)
The search_goodreads()
function allows you to search for books on Goodreads based on various criteria.
The code below searches for books that include the term “parenting” in the title and returned 10 books sorted by readers’ ratings
parent_df <- search_goodreads(search_term = "parenting", search_in = "title", num_books = 10, sort_by = "ratings")
summary(parent_df)
## title author book_id
## Length:10 Length:10 Length:10
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
## url ratings
## Length:10 Min. : 8427
## Class :character 1st Qu.:11744
## Mode :character Median :13662
## Mean :19757
## 3rd Qu.:13784
## Max. :69591
You can also search author’s name:
search_goodreads(search_term = "J.K. Rowling", search_in = "author", num_books = 5, sort_by = "ratings")
The search_goodreads()
function includes a sort_by
that sorts the results either by ratings
or published_year
:
search_goodreads(search_term = "J.K. Rowling", search_in = "author", num_books = 5, sort_by = "published_year")
After the books are found, save their IDs to a text file. These IDs are used for extracting book metadata and reviews:
get_book_ids(input_data = parent_df, file_name = "parent_books.txt") #the book IDs are now stored in a text file named “parent_books”
Book metadata can then be scraped:
parent_bookinfo <- scrape_books(book_ids_path = "parent_books.txt", use_parallel = FALSE)
To speed up the scraping process:
*Turn on the parallel process: use_parallel = TRUE
*Specify the number of cores for the parallel process (e.g., `num_cores = 8)
parent_bookreviews <- scrape_reviews(book_ids_path = "parent_books.txt", num_reviews = 10, use_parallel = FALSE) #users can also turn on parallel process to speed up the process
The analyze_sentiment()
function calculates the sentiment score of each review based on the lexicon chosen by the user. Available options for lexicon are afinn
, bing
, and nrc
. Basic negation scope detection was implemented (e.g., not happy is labeled as negative emotion and is assigned with a negative score).
sentiment_results <- analyze_sentiment(parent_bookreviews, lexicon = "afinn")
The average_book_sentiment()
function calculates the average sentiment score for each book.
ave_sentiment <- average_book_sentiment(sentiment_results)
summary(ave_sentiment)
## book_id avg_sentiment
## Length:10 Min. : 4.40
## Class :character 1st Qu.: 7.25
## Mode :character Median :12.86
## Mean :12.95
## 3rd Qu.:14.65
## Max. :27.30
The sentiment scores can be plotted as a histogram:
sentiment_histogram(sentiment_results)
Or a trend of average sentiment score over time:
sentiment_trend(sentiment_results, time_period = "year")
Apply topic modeling to the reviews data:
reviews_topic <- model_topics(parent_bookreviews, num_topics = 3, num_terms = 10, english_only = TRUE)
## Topic 1:
## parent, children, need, one, way, good, get, work, dont, give
##
## Topic 2:
## parent, child, book, emot, feel, help, also, can, children, use
##
## Topic 3:
## book, just, kid, think, read, like, time, say, realli, much
Plot the top terms by topic:
plot_topic_terms(reviews_topic)
Create a word cloud for each topic:
gen_topic_clouds(reviews_topic)
Topic 1:
Topic 2:
Topic 3: