content-based-rec.py

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import linear_kernel

# import data
books = pd.read_json("data/goodreads_books_poetry.json", lines = True)
reviews = pd.read_json("data/goodreads_reviews_poetry_cleaned.json", lines = True)

# merge both books and reviews based on book_id
books_and_reviews = pd.merge(reviews, books, on=['book_id'])

# filter out only the english books
books_and_reviews = books_and_reviews.loc[books_and_reviews['language_code'].isin(['en-US', 'eng', 'en-GB', 'en-CA'])]
# make one entry per book
books_and_reviews = books_and_reviews.groupby(['book_id', 'title', 'description'], as_index=False).agg(' '.join)
# add extra column of only text (title, description, and review text)
books_and_reviews['text_data'] = books_and_reviews['title'] + books_and_reviews['description'] + books_and_reviews['review_text'] 

# initialize vectorizer from sklearn and remove stopwords
tf_idf = TfidfVectorizer(stop_words='english')
# generate vectors for text data
vectors = tf_idf.fit_transform(books_and_reviews['text_data'])
# calculate cosine similarities (matrix)
cosine_sim_matrix = linear_kernel(vectors, vectors)

# decision module
indices = pd.Series(books_and_reviews.index, index=books_and_reviews['book_id']).drop_duplicates()
#function to get recommendation based on a book_id
def recommend_book(book_id):
    # get the index of the book we're referencing/comparing to for recommendations
    index = indices[book_id]
    
    # get pairwise cosine similarities
    cosine_sim = list(enumerate(cosine_sim_matrix[index]))
    
    # sort on cosine similarity scores (descending order)
    cosine_sim = sorted(cosine_sim, key=lambda x: x[1], reverse=True)
    
    # get the top 10 books
    cosine_sim = cosine_sim[1:11]
    top_10 = [score[0] for score in cosine_sim]
    return top_10

#select a random book to use and generate recommendations on
random_book = books_and_reviews.sample()
book_selected = random_book['book_id'].values[0]
print('=================book selected=================')
print(random_book['title'].values[0])

print('=================testing - generating recommendations=================')
# pass the book_id to recommend_book to get the top 10 book recommendations
top_10 = recommend_book(book_selected)
for book_index in top_10:
    book = books_and_reviews.iloc[book_index]
    id = book.book_id
    print(book.title)


# USE CASE 1
book_title_1 = "Love Poems"
book_1 = books_and_reviews[books_and_reviews['title'] == book_title_1]
book_id_1 = book_1.iloc[0].book_id if not book_1.empty else 0
print(book_id_1 if book_id_1 != 0 else "PLEASE ENTER ANOTHER BOOK TITLE")

print('=================use case 1: book selected=================')
print(book_title_1)
print('=================use case 1: testing - generating recommendations=================')
# pass the book_id to recommend_book to get the top 10 book recommendations
top_10_1 = recommend_book(book_id_1)
for book_index in top_10_1:
    book = books_and_reviews.iloc[book_index]
    id = book.book_id
    print(book.title)


# USE CASE 2
book_title_2 = "Salt"
book_2 = books_and_reviews[books_and_reviews['title'] == book_title_2]
book_id_2 = book_2.iloc[0].book_id if not book_2.empty else 0
print(book_id_2 if book_id_2 != 0 else "PLEASE ENTER ANOTHER BOOK TITLE")
print('=================use case 2: book selected=================')
print(book_title_2)
print('=================use case 2: testing - generating recommendations=================')
# pass the book_id to recommend_book to get the top 10 book recommendations
top_10_2 = recommend_book(book_id_2)
for book_index in top_10_2:
    book = books_and_reviews.iloc[book_index]
    id = book.book_id
    print(book.title)


# USE CASE 3 - book not in dataset (you can use this to test - just replace the book title with whatever poetry book)
''' For this case, we will select a Harry Potter book which is clearly not a poetry book! 
This will show "PLEASE ENTER ANOTHER BOOK TITLE" - where you can replace this with your desired 
poetry book for testing. Once you have a valid book, you will get back the book id and you can uncomment 
the code block below.
'''
book_title_3 = "Harry Potter and the Sorcerer's Stone"
book_3 = books_and_reviews[books_and_reviews['title'] == book_title_3]
book_id_3 = book_3.iloc[0].book_id if not book_3.empty else 0
print(book_id_3 if book_id_3 != 0 else "PLEASE ENTER ANOTHER BOOK TITLE")
# print('=================use case 3: book selected=================')
# print(book_title_3)
# print('=================use case 3: testing - generating recommendations=================')
# # pass the book_id to recommend_book to get the top 10 book recommendations
# top_10_3 = recommend_book(book_id_3)
# for book_index in top_10_3:
#     book = books_and_reviews.iloc[book_index]
#     id = book.book_id
#     print(book.title)