Dataset Card for Amawal.net-Dataset

This dataset compiles the crowdsourced lexicon and linguistic entries from the legacy platform Amawal.net. Following the closure of the original website, this repository provides an archive of the community's multi-dialectal contributions to preserve its linguistic value and make it accessible for natural language processing (NLP), lexicography, and digital humanities applications.

Dataset Details

Dataset Description

The Amawal.net Dataset is a corpus consisting of over 10,000 forum-style entries contributed by internet users up until its data cutoff in 2011. The structural format mirrors conversational forum posts: each record contains a core word as the title, followed by contextual content detailing the word's description, semantic meaning, regional translations, script transliterations, and dialectal variations.

Due to its organic, crowdsourced origin, the completeness of fields varies per entry, offering a highly realistic cross-section of early digital pan-Amazigh lexical compilation.

• Original Data Creators: The crowdsourced user community of Amawal.net
• Language(s) (NLP): Broad regional variants of Amazigh (including Standard Moroccan Amazigh zgh, Kabyle kab, Tashelhit shi, Central Atlas Tamazight tzm, Rifian rif) mapped with parallel translations or descriptions in French (fr), Spanish (es), English (en) and Arabic (ar).
• License:

Dataset Sources

• Original Platform: Amawal.net (Defunct/Offline)

Uses

Direct Use

• Lexicon & Dictionary Extraction: Mining structured terms, neologisms, and dialectal synonyms to build specialized electronic glossaries.
• Language Modeling & Tokenization: Training small language models (SLMs) and morphological analyzers on descriptive multilingual data across multiple scripts.
• Cross-Dialectal Linguistic Mapping: Studying comparative vocabulary adjustments between northern variants (like Kabyle or Rifian) and southern variants (like Tashelhit).

Out-of-Scope Use

This dataset is an archived snapshot from 2011. It should not be used as an absolute reference manual for official linguistic standardization frameworks (such as IRCAM standard spelling guidelines) without pre-processing, as entries reflect diverse community conventions of that era.

Dataset Structure

The dataset drops rigid tabular constraints to follow a forum-like document layout where entries populate two core string entities:

• title: The target word or term being defined.
• content: The full text block of the post containing meanings, script transliterations, and descriptive translation vectors depending on what the users provided.

Dataset Creation

Curation Rationale

Amawal.net represents one of the earliest and most extensive community-driven efforts to digitize the Amazigh vocabulary on the internet. Because the website is permanently offline, this dataset acts as a digital preservation repository to secure this collaborative heritage for low-resource computational research before it is lost.

Source Data

Data Collection and Processing

Who are the source data producers?

The dataset was generated between the early 2000s and 2011 by a diverse, multilingual network of Amazigh speakers, teachers, activists, and language learners collaborating on the web.

Personal and Sensitive Information

Bias, Risks, and Limitations

• Varying Completeness: Some entries consist of comprehensive examples and multi-language translations, while others contain only simple translation equivalents.
• Historical Orthography: Content created prior to 2011 may feature localized, inconsistent script conventions or custom character variations that predated widespread digital standardization.

Glossary

• Amawal: The lexical term for "Dictionary" or "Glossary" within the Amazigh language.
• Crowdsourced Lexicography: Collaborative creation of dictionaries where decentralized public users write and submit definitions directly.

Dataset Card Authors

• Contact: abdelhaque.idali@gmail.com