Kdramarama is a web application for k-drama recommendations. Recommendations are based on my personal reviews of 150+ k-dramas. View the app here: https://k-drama-rama.com

• Get recommendations by selecting a k-drama from a list of titles
• View details of recommended dramas including genres, ratings, and descriptions

• Frontend: React, TailwindCSS
• Backend: FastAPI
• Scraper: BeautifulSoup, requests, selenium
• NLP: NLTK, gensim, spacy, textblob
• Recommendations: scikit-learn
• Deployment:
  • Backend → Heroku
  • Frontend → Vercel

Data on dramas (titles, images, ratings, genres, etc.) is sourced from MyDramaList. My personal watchlist and reviews are the primary text source for feature engineering (more details below).

I use my personal reviews of 150+ kdramas as the basis for the recommendation engine. Steps include:

• Text Cleaning & Preprocessing: Remove punctuation, convert to lowercase, tokenize text, remove stopwords, and lemmatize words for analysis.
• Word & Phrase Frequency Analysis: I use CountVectorizer and TextBlob to extract common words and phrases in my drama reviews, and visualize the results with word clouds.
• Semantic Similarity Analysis: I then use Word2Vec embeddings to measure similarity between extracted phrases based on cosine similarity in order to consolidate similar phrases and refine a list of common themes to provide a starting point for feature engineering.
• Sentiment Analysis: I use NLTK's sentiment library to generate sentiment polarity scores for each review, as well as context-specific polarity scores for specific key phrases found in reviews (these phrases are gleaned in the previous step).
• Feature Engineering: I use review and phrase-specific polarity scores as features. Additionally, I also create binary features based on key words and phrases found in reviews.
• Recommendation Model: My recommendation model computes cosine similary for all drama titles in order to generate the N most similar dramas to a selected title.

git clone https://github.com/dsherbini/kdrama-recommendations
cd kdrama-recommendations

• Python 3.11+

cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

uvicorn main:app --reload

The API should be available at http://localhost:8000.

cd frontend
npm install
npm start

The frontend will run at http://localhost:3000.

Create a .env file in the frontend/ directory:

REACT_APP_API_URL=http://localhost:8000