🎶 Bollywood Songs & Albums Metadata Dataset (1931–2025)

A comprehensive, structured, and reproducible metadata dataset of Bollywood film albums and songs spanning over 90 years (1931–2025).

This dataset is designed for:

• 🎧 Music information retrieval
• 📊 Data analysis & visualization
• 🤖 Machine learning & recommendation systems
• 🧠 Digital humanities & cultural research
• 🏗️ Scalable music databases

---

📌 Dataset Overview

Category	Count
🎞️ Albums	12,673
🎵 Songs	57,005
📅 Year Range	1931 – 2025
📁 Format	CSV
🔑 Identifiers	Deterministic UUIDv5
🔁 De-duplication	Yes (content-based)

---

📝 Data Dictionary

Detailed schema for the CSV files found in data/raw/.

1. Albums Schema (albums.csv)

Each row represents a unique film soundtrack or album.

Column Name	Type	Description	Example
album_uuid	string	Primary Key. Deterministic UUIDv5 generated from title + year.	a3f9...
album_title	string	Normalized title of the film/album.	Sholay
album_year	int	Release year of the album.	1975
music_director	string	Primary composer(s) of the album.	R.D. Burman
label	string	Music label (e.g., T-Series, Saregama), if available.	Polydor

2. Songs Schema (songs.csv)

Each row represents a unique track within an album.

Column Name	Type	Description	Example
song_uuid	string	Primary Key. Deterministic UUIDv5 generated from album + track + title.	c1b2...
album_uuid	string	Foreign Key. Links to the parent Album.	a3f9...
track_number	int	Sequential order of the song in the album.	1
song_title	string	Title of the track.	Mehbooba Mehbooba
singers	string	Comma-separated list of singers.	R.D. Burman
1 youtube_url	string	Direct link to the official music video or audio.	https://youtu.be/...
3 yt_music_urls	string	Direct link to the official music track	https://music.youtube.com/watch?v=.....

🗂️ Dataset Structure

data/
├── raw/
│   ├── albums/
│   │   ├── albums_1931_1944.csv
│   │   ├── albums_1945_1954.csv
│   │   ├── ...
│   │   └── albums_2015_2025.csv
│   └── songs/
│       ├── songs_1931_1944.csv
│       ├── songs_1945_1954_completed.csv
│       ├── ...
│       └── songs_2015_2025.csv

📊 Verified Record Counts

💿 Albums (12,673 total)

Period	Albums
1931–1944	387
1945–1954	1,066
1955–1964	1,058
1965–1974	1,079
1975–1984	1,173
1985–1994	1,245
1995–2004	1,127
2005–2014	1,551
2015–2025	3,987

---

🎵 Songs (57,005 total)

Period	Songs
1931–1944	1,606
1945–1954	6,507
1955–1964	6,757
1965–1974	5,411
1975–1984	5,950
1985–1994	6,862
1995–2004	6,248
2005–2014	8,191
2015–2025	9,473

---

🧬 Deterministic UUID Design (UUIDv5)

This dataset uses UUID version 5 to ensure stable, repeatable, and collision-free identifiers across re-scrapes and future updates.

---

💿 Album UUID Generation

Each album UUID is generated from a normalized, human-readable album identity.

Inputs

• album_title
• album_year

unique_album_string = f"{album_title}_{album_year}".lower().strip()
album_uuid = uuid.uuid5(uuid.NAMESPACE_DNS, unique_album_string)

Why this works

• Same album title + year → same UUID forever
• Prevents duplicate albums across re-scrapes
• Stable primary key for joins, updates, and merges
• Human-explainable identity source

---

🎵 Song UUID Generation

Each song UUID is generated relative to its parent album, ensuring correct hierarchy and uniqueness.

Inputs

• album_uuid (parent)
• track_number
• song_title

unique_song_string = f"{album_uuid}_{track_number}_{song_title}".lower().strip()
song_uuid = uuid.uuid5(uuid.NAMESPACE_DNS, unique_song_string)

Why this works

• Same album + track + title → same song UUID
• Allows identical song titles across different albums
• Enforces album → song referential integrity
• Ideal for relational databases and graph models

---

🧠 Relational Model

Album (album_uuid)
 ├── album_title
 ├── album_year
 ├── music_director
 └── label
        │
        └── Song (song_uuid)
            ├── track_number
            ├── song_title
            ├── singers
            └── youtube_url

---

🧪 Example Use Cases

• 🎼 Music recommendation engines
• 📈 Trend analysis across decades
• 🎤 Singer & composer network graphs
• 🤖 ML training datasets
• 🗃️ Music archival systems

---

📦 Kaggle & HuggingFace Ready

• Flat CSV files
• Stable UUID primary keys
• Clear schema
• Deterministic regeneration
• Version-friendly structure

Perfect for:

• Kaggle Datasets
• HuggingFace datasets
• BigQuery / DuckDB / Postgres imports

---

📖 Cite This Dataset

If you use this dataset in research, projects, or publications, please cite it as:

@dataset{bollywood_metadata_1931_2025,
  title   = {Bollywood Songs and Albums Metadata Dataset (1931--2025)},
  author  = {Asacker},
  year    = {2026},
  url     = {https://github.com/MrAsacker/bollywood-songs-metadata},
  note    = {12,673 albums and 57,005 songs with deterministic UUIDv5 identifiers}
}

---

⚖️ License

This project and dataset are released under the MIT License.

You are free to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the dataset and code, subject to the terms of the MIT License.

See the LICENSE file for full details.

If this dataset helped you, consider giving the repo a ⭐ — it really helps.