jamalinu/amazigh-nlp-spacy

# 🌍 Amazigh/Berber NLP Pipeline with spaCy > First open-source NLP pipeline for Tamazight (Amazigh/Berber language) > built by a native speaker with formal linguistics training. ## 🎯 Why this project matters Tamazight is spoken by **30+ million people** across North Africa, yet has virtually zero computational linguistics resources. This project addresses that gap by combining: - Native speaker linguistic knowledge (Tamazight, Arabic, Spanish, French, Catalan) - Formal philology and linguistics training - Modern NLP tools (spaCy, scikit-learn, Python) --- ## 🚀 Live Demo **👉 Try it here — amazigh-nlp-demo on Hugging Face** The demo includes three tools running live: | Tool | Description | |------|-------------| | 🔎 Named Entity Recognizer | Detects people (PER) and locations (LOC) | | 🎭 Sentiment Analyzer | Classifies comments as Positive / Negative / Neutral with confidence % | | 🏔️ Tarifit Morphology Analyzer | POS tagging, clitic tokenization and morpheme segmentation | | 🎙️ TTS Text Normalizer | Converts informal Tarifit (Chat-Arabic numerals, digraphs) to phonetic symbols | --- ## 🔬 Project Modules | # | Module | Description | Status | |---|--------|-------------|--------| | 01 | Tokenizer | Custom Tamazight tokenizer with clitic handling | ✅ Done | | 02 | NER Data Prep | DocBin annotation pipeline | ✅ Done | | 03 | NER Training | Supervised NER model (100% F1) | ✅ Done | | 04 | Sentiment Analysis | Positive / Negative / Neutral classifier | ✅ Done | | 05 | TTS Corpus | Phonetically Balanced Corpus for Tarifit TTS | ✅ Done | | 06 | Morphology | Full morphological pipeline for Tarifit | ✅ Done | | 07 | TTS Frontend | Linguistic text normalization for Tarifit TTS | ✅ Done | | 08 | Corpus | Annotated sentences dataset | 🔄 In progress | --- ## 📓 Notebooks ### 01. Custom NLP Pipeline for Low-Resource Languages Builds a spaCy tokenizer from scratch for Tamazight. Handles clitic segmentation and Tifinagh character normalization. --- ### 02. NER Data Prep …