dedogospod/chalga-lyrics-classified

Balkan Chalga Lyrics Dataset (Classified)

A dataset of 7,132 classified Balkan chalga/pop song lyrics scraped from lyricsmusic.eu. Each entry contains the original URL, title, artist name, full lyrics text, a cleaned/normalized artist name (classified_artist), and an automatically detected language code.

Dataset Statistics

Metric	Value
Total entries	7,132
Languages represented	15
Primary language	Bulgarian (bg) — 88.8%
Train / Val / Test split	80% / 10% / 10%

Language Breakdown

| Language Code | Count | Percentage | | bg | 6,335 | 88.8% | | sr | 459 | 6.4% | | el | 181 | 2.5% | | mk | 36 | 0.5% | | tr | 34 | 0.5% | | hr | 33 | 0.5% | | ru | 16 | 0.2% | | ko | 11 | 0.2% | | es | 11 | 0.2% | | ro | 6 | 0.1% | | ... | 3 | 0.0% | | fr | 2 | 0.0% | | sq | 2 | 0.0% | | he | 2 | 0.0% | | zh | 1 | 0.0% |

Languages

• bg — Bulgarian (Български)
• sr — Serbian (Српски)
• el — Greek (Ελληνικά)
• mk — Macedonian (Македонски)
• tr — Turkish (Türkçe)
• hr — Croatian (Hrvatski)
• ru — Russian (Русский)
• Plus small amounts of Korean, Spanish, Romanian, French, Albanian, Hebrew, and Chinese

Schema

Column	Type	Description
url	string	Source URL from lyricsmusic.eu
title	string	Song title (as scraped)
artist	string	Artist name (as scraped)
lyrics	string	Full song lyrics text
classified_artist	string	Cleaned/normalized artist name
language	string	ISO 639-1 language code

Collection & Classification Methodology

1. Scraping: Lyrics were scraped from lyricsmusic.eu, a multilingual lyrics website covering Balkan and Eastern European music.
2. Language Classification: Each entry was classified using Qwen 3.6 27B via the LM Studio API, which identified the primary language of each song's lyrics.
3. Filtering: English-language songs were filtered out separately (83 entries discarded) to focus on non-English Balkan music content.

Splits

The dataset is split into three subsets:

Split	Entries	Percentage
Train	5,705	80%
Validation	713	10%
Test	714	10%

Splits were generated with seed=42 for reproducibility.

Usage

from datasets import load_dataset

# Load the full dataset
dataset = load_dataset("dedogospod/chalga-lyrics-classified")

# Access splits
train_ds = dataset["train"]
val_ds = dataset["validation"]
test_ds = dataset["test"]

# Example entry
print(train_ds[0])

License & Copyright Notice

This dataset contains copyrighted song lyrics. The lyrics are the intellectual property of their respective artists, composers, and publishers. This dataset is provided for research and educational purposes only. Commercial use or redistribution of the lyrics content may violate copyright law in your jurisdiction. Please respect the rights of music creators and publishers.

Citation

If you use this dataset in your research, please cite it as:

@dataset{dedogospod_chalga_lyrics_2025,
  author       = {Asen Mitrev},
  title        = {Balkan Chalga Lyrics Dataset (Classified)},
  year         = 2025,
  url          = {https://huggingface.co/datasets/dedogospod/chalga-lyrics-classified},
  note         = {Scraped from lyricsmusic.eu, classified with Qwen 3.6 27B}
}

Author

Created by Asen Mitrev for research on multilingual Balkan music classification and NLP tasks.