Task :Recognising book genre- based on goodreads dataset. We will use different features-including text features(description and title), image features(Book covers) and other numerical features(such as number of pages, ratings and more) in order to recognise a book genre.

(updates were made from danielaneuralx which is my working github but its all mine.)

• eda_goodreads.ipynb (include exploration of dataset)
• inference_goodreads.ipynb (learning models)

Report: 701projectreport_bookgenre.pdf

Project structure:

• root
  • data
    • images-source(Directory-don't delete -this holds dataset images. It Exists in github)
      • 1.jpg (image file)
      • 2.jpg (image file)
      • ..(230K images)
      • books_images_names.csv
      • goodreads_imagestxt.txt
    • images-train (Directory-This directory and content will be created)
    • images-val(Directory-This directory and content will be created)
    • images-test(Directory-This directory and content will be created)
    • goodreads_books_eng_f1.csv (Dont delete-This is the first dataset csv)
    • goodreads_books_eng_f2.csv (Dont delete-This is the second dataset csv)
  • goodreads (package)
    • init.py
    • baseline.py
    • conv_goodreads.py
    • custom_nn_with_embeddings
    • results_utils.py
    • utils.py
  • configuration.yml (Very important -dont delete)= this holds hyperparameters configuration and general parameters
  • eda_goodreads.ipynb
  • inference_goodreads.ipynb

Needs packages:

• tensorflow
• matplotlib
• sckit-learn
• os
• pyyaml
• numpy
• pandas
• tesnorflow-addons (for F1 metric)
• nltk
• gensim
• spacy
• nltk
• pickle
• shutil

should run : python -m spacy download en_core_web_md

Please for any problem or question-find me at [email protected]