🎵 Music Recommender System

A lightweight content-based music recommendation app built with Python.
It uses TF-IDF vectorization and cosine similarity on song lyrics to suggest similar tracks.

🚀 Deployed with Streamlit for an intuitive, interactive experience.
🎧 Get instant music recommendations with Spotify album covers integrated.

---

🎥 Demo

Features

• 🎼 Content-based Recommendations using TF-IDF vectorization on lyrics/text features + Cosine Similarity to suggest similar tracks.
• ⚡ Lightweight & Git-friendly: avoids committing large files (committed files are computed dynamically at runtime)
• 🎨 Streamlit UI: clean design with Spotify album covers
• 🔍 Interactive Search: choose songs from dropdown

---

Project Structure

├── preprocess.py    # Cleans dataset, builds TF-IDF & cosine similarity
├── recommend.py     # Loads data & computes song recommendations
├── main.py          # Streamlit app interface (UI + Spotify album covers)
├── requirements.txt # Python dependencies
├── README.md        # Documentation

---

Dataset

This project uses the Spotify Million Song Dataset (Lyrics Data) from Kaggle.

• The dataset contains song lyrics and metadata (artist, track, link).
• In this project:
  • The text (lyrics) column is cleaned and preprocessed with NLTK.
  • A TF-IDF matrix is built from the cleaned text.
  • Cosine similarity is calculated to recommend top-N similar songs.

⚠️ Note: Due to size, the dataset itself is not included in this repository. Please download it from Kaggle if you want to preprocess it yourself.

---

📊 Data Preprocessing & EDA

Before building the recommender, the raw dataset required preprocessing:

• 🔤 Removed links, numbers & special characters
• ✂️ Converted text to lowercase, tokenized, and removed stopwords
• 📐 Built a TF-IDF matrix (max 5000 features)
• 🔗 Computed cosine similarity between song vectors

👉 For detailed preprocessing and exploratory data analysis, see the notebook:
Music_Recommendation_System.ipynb

---

🚀 Quick Start

1. Clone the Repository

   git clone https://github.com/Sonalikasingh17/Music_Recommender_System.git
   cd Music_Recommender_System

2. Install Dependencies

   pip install -r requirements.txt

3. Download Dataset

   Download the dataset from Kaggle:

   👉 Spotify Million Song Dataset

4. Run Preprocessing

   python preprocess.py

• This script cleans the dataset and create these files in project folder:
  • df_cleaned.pkl
  • tfidf_matrix.pkl
  • cosine_sim.pkl

5. Launch the Streamlit app

   streamlit run main.py

   Open the local URL in your browser to explore the app.

---

⚡Tips & Customization

• Autocomplete Search Bar: Enhance the select box with st.selectbox(..., help="Type to search...") or use st.text_input() + fuzzy matching for better UX.
• Custom Styling: Modify CSS in main.py to refine recommendation cards (colors, fonts, spacing).
• Data Customization: Replace df['text'] preprocessing logic to work with lyrics, metadata, or genre features.

---

🙏 Acknowledgements

This project was built with ❤️ using scikit-learn, pandas, NLTK, and Streamlit,
inspired by common TF-IDF + cosine similarity recommender patterns for content-based filtering.
Special thanks to the Spotify Million Song Dataset.

---

✨ Enjoy exploring music recommendations with your interactive, lightweight app! ✨