abdelhaqueidali/Classical-Poetry-Dataset

Classical Southern Moroccan Amazigh Music Lyrics Dataset

This dataset contains scraped titles, YouTube video links, and descriptions featuring lyrics for classical Southern Amazigh (Tashelhit) music. The collection spans over 400 songs, serving as a valuable text corpus for regional linguistic study, poetic analysis, and natural language processing (NLP) tasks in low-resource variants.

Orthography & Linguistic Note

The orthography captured in this dataset represents spoken Amazigh (Tashelhit) as transcribed natively for the source channel. While it may not strictly follow standardized academic conventions, it provides an authentic reflection of spoken Tashelhit transcribed into both Tifinagh and Arabic scripts, which are both present for the majority of the entries.

---

Acknowledgments & Provenance

The lyrics in this dataset exist thanks to the dedicated transcription work of Mohamed Ouadi, who manually compiled and wrote these lyrics to preserve old Tashelhit audio-visual poetry.

---

Dataset Structure

This dataset is organized into two primary files that reflect a multi-layered curation process for regional poetry and lyrics:

• channel_final.json (Main Viewer Dataset): This is the comprehensive, unified production database. It preserves the complete video metadata and descriptions while seamlessly appending the corresponding text scraped from linked community posts.
• posts.json (Extraction Split): This file contains records where a video directly links to a dedicated community post. While the core video entries duplicate what is found in the main channel file, these entries are uniquely enriched with targeted post metadata and the text of the post itself (content) which is either a completion, or alternative script transliteration, or repost of the same lyrics found on the description, and might as well be another lyrics of the same artist.

Contextual Variants in Community Posts

The text retrieved from the community posts and merged into these files provides critical linguistic variations that complete or expand upon the standard video details. The crawled post content generally fulfills one of four distinct criteria:

1. Alternative Script Variations: Provides the exact poem rendered in an alternative orthography (such as the Amazigh alphabet) to complement the Arabic script found in some videos descriptions.
2. Lyrical Completion: Supplies the full text or missing stanzas of a poem that was truncated in the video description due to character limits.
3. Artist/Poet Extensions: Features an entirely different poem or continuation by the same artist that contextualizes the track.
4. Linguistic Verification: Serves as an exact, verified repost of the lyrics to ensure textual integrity and orthographic consistency across the library.

Key Schema Fields

Entries across the dataset maintain the following structure where applicable:

• url: The primary YouTube video link destination.
• title: The native title of the track/video resource.
• description: The raw description text from the video (often containing Arabic script transcriptions).
• post_url: The permalink to the associated YouTube community post (null if none exists).
• attached_video_url: Video links mapped inside the post itself.
• content: Cleaned, un-truncated text extracted directly from the post (stripped of URLs and hashtags).

---

Use Cases

• NLP & Text Mining: alignment, and language modeling for low-resource Amazigh languages.
• Linguistic & Literary Analysis: Studying the poetic structure, metaphors, and vocabulary of classical Southern Amazigh music (Amarg / Rways).
• Cultural Preservation: Digital archiving of oral poetry and heritage.

Souce

https://youtube.com/@mohamedouadi