NLP Research
Darija NLP Pipeline
From speech to sentiment: NLP for Moroccan Darija.
- Year
- 2024 — 2025
- Role
- Research & engineering
- Type
- NLP Research
Overview
Two complementary tools forming a Darija language pipeline: an audio/video-to-SRT transcriber for Arabic and Darija (Wit.ai + voice-activity segmentation), and a sentiment classifier for social media comments (scraping → preprocessing → classification served over a REST API). Part of ongoing research interest in under-resourced dialects.
Highlights
- VAD-segmented transcription pipeline for Arabic/Darija media
- Sentiment classification on scraped Facebook/YouTube comments
- Foundation for darija-focused NLP research
Stack
- Python
- scikit-learn
- Wit.ai
- webrtcvad
- ffmpeg
- Flask
