NextDrama is a content-based recommendation system for Asian dramas, movies, specials, and TV shows. It leverages deep semantic similarity using Sentence Transformers to suggest similar titles based on what you love — no user history required.

🌐 Live Demo: nextdrama.streamlit.app

• 🤖 Semantic similarity with all-mpnet-base-v2 from Sentence Transformers
• 🧠 Content-based recommendations — no collaborative filtering needed
• 🎛️ Filter by genre, country, type, rating, tags, and release year
• 🖼️ Interactive UI with responsive grid layouts
• 🪪 Card-based display with full metadata: cast, synopsis, creators, etc.
• ⚡ Fast cosine similarity search over 20,000+ media embeddings

The system is powered by the Asian Drama Dataset on Kaggle, which includes:

• Over 20,000 titles with rich metadata
• Fields like title, synopsis, cast, genres, tags, score, country, and network
• Preprocessed and ready for NLP tasks or recommender systems

• Embedding Model: all-mpnet-base-v2
• Input Text: A combined "soup" of title, synopsis, genres, tags, cast, and directors
• Similarity Metric: Cosine similarity via sentence_transformers.util.cos_sim
• Inference: Fast top-k search with PyTorch tensors
• Evaluation: 80,000+ human-labeled positive recommendation pairs for recall@K evaluation

# Clone the repository
git clone https://github.com/lakhindarpal/NextDrama.git
cd NextDrama

# Install dependencies
pip install -r requirements.txt

# Run the app
streamlit run app.py

You can also manually load the precomputed data:

import joblib

media_df = joblib.load("media_df.pkl")
embeddings = joblib.load("embeddings.pkl")

📌 Click on a title card to reveal full information including synopsis, cast, tags, genres and more.

📱 Mobile	💻 Desktop

The current model uses cosine similarity over transformer-based embeddings. Evaluation can be performed offline using the 80K labeled recommendation pairs and metrics like Recall@K.

Currently live on Streamlit Cloud: 👉 nextdrama.streamlit.app

You can also deploy on:

• 🐳 Docker
• 🤗 Hugging Face Spaces
• 🔧 Self-hosting or private servers

• Streamlit — frontend framework
• SentenceTransformers — semantic embeddings
• PyTorch — tensor computation
• Joblib — model/data persistence
• Pandas — data handling

This project is licensed under the MIT License.

• Sentence Transformers
• Streamlit
• Hugging Face
• Kaggle Dataset

Found a bug? Have a suggestion? Open an issue or drop feedback for improvements!