README.md

metadata

license: apache-2.0
dataset_info:
  features:
    - name: text
      dtype: string
    - name: song_name
      dtype: string
    - name: artist_name
      dtype: string
  splits:
    - name: train
      num_bytes: 4563835
      num_examples: 1451
  download_size: 2464487
  dataset_size: 4563835
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - text-generation
  - text2text-generation
language:
  - en
tags:
  - code
  - music
pretty_name: RapLyric
size_categories:
  - 1K<n<10K

---

Dataset Card for RapLyrics

Dataset Description

Dataset Summary

The "RapLyrics" dataset contains ~1.4k lyrics from rap songs collected using the Genius API. The dataset includes metadata such as artist names, song titles, and lyrics. The lyrics were crawled and processed to ensure quality and consistency. This dataset is intended for research and development in natural language processing, music analysis, and related fields.

Data Cleaning

The dataset underwent several cleaning steps to ensure the quality of the lyrics:

1. Content Removal:

   • Removed content before the word "Lyrics".
   • Removed names after tags like [Intro], [Verse], etc.
   • Removed duplicate sections of lyrics, keeping only the first occurrence of each section.

2. Repetition Handling:

   • Detected and removed overall repetition by comparing the first and second halves of the lyrics.
   • Removed duplicate sections based on tags (e.g., [Chorus], [Verse 1]).

3. Line Breaks and Empty Lines:

   • Preserved line breaks before tags.
   • Removed empty lines to improve readability.

4. Filtering Criteria:

   • Excluded lyrics with fewer than 1024 characters.
   • Excluded lyrics with fewer than 100 unique words.

5. Regex Cleaning:

   • Removed irrelevant content such as advertisements (e.g., "See ... Get tickets as low as $X").
   • Removed text inside brackets (e.g., [Intro], (Chorus)).
   • Removed punctuation while preserving language characters.

Notes/Limitations

• The dataset may contain outliers in length or content due to imperfections in the Genius search and crawling process.

Dataset Structure

The dataset is organized as follows:

• Each row corresponds to a song, with its song name, song artist and lyrics recorded.