abdelhaqueidali/Lexicon-Matrix-Dglai

Amazigh Lexical Dataset

This dataset is an expanded, structured NLP instruction-tuning and grammatical matrix for Standard Moroccan Amazigh (Tamazight). It converts raw lexicographical data into clean declarative descriptions and rich question-answering (QA) frames covering linguistic morphology, parts of speech, and multi-way translations between Amazigh, English, French, and Arabic.

Dataset Structure

The dataset is formatted as a JSON object where each key is a unique lexical identifier (id_lexie). Every entry contains two primary components: metadata (pristine raw fields) and generated_frames (natural prose generation and QA tasks).

Field Mapping Reference

The following table maps the raw dictionary CSV headers, their linguistic acronyms, and the properties utilized during generation:

CSV Field Name	Linguistic Term / Feature	English Value / Processing Role
id_lexie	Lexical Identifier	Core Dataset Row ID Key
lexie	Base Lexeme	w (Amazigh Alphabet Orthography)
api	Alphabet Phonétique International	ap (Latin System / IPA Transliteration)
cg	Catégorie Grammaticale	cg_en (Clean English Category Tag)
eadata	État Annexe Singular	sc (Singular Construct State)
pldata	Pluriel Libre	pf (Plural Free State)
peadata	Pluriel État Annexe	pc (Plural Construct State)
fel	Féminin Singulier Libre	fs (Feminine Singular Free Form)
fea	Féminin Singulier État Annexe	fsc (Feminine Singular Construct State)
fpel	Féminin Pluriel Libre	fpf (Feminine Plural Free Form)
fpea	Féminin Pluriel État Annexe	fpc (Feminine Plural Construct State)
variante	Variant	vt (Alternate Spelling / Dialect Variant)
acc	Accompli	va_acc (Accomplished Aspect Verb Form)
acc_neg	Accompli Négatif	va_acc_neg (Negative Accomplished Aspect Verb Form)
inacc	Inaccompli	va_inacc (Unaccomplished Aspect Verb Form)
senseng_joined	English Meaning	em_en (English Meaning / Definition Context)
sensfr_joined	Sens Français	em_fr (French Meaning / Definition Context)
sensar_joined	Meaning Arabic	em_ar (Arabic Meaning / Definition Context)

---

Source & Acknowledgements

The foundational lexical entries, grammatical properties, and core translations driving this dataset engine are sourced directly from the IRCAM DGLAI platform:

• Raw Source Dictionary: abdelhaqueidali/ircam-dglai-dataset

• IF YOU NOTICE ANY ISSUE OR INCONSISTENCY, PLEASE CONTACT ME:

• abdelhaqueidali@gmail.com