Logo Lanfrica

mohammel-debug/loan-words-darija

Domaine:

natural language processing
Créateur:
moh
Hôte:
# Loanword Extraction in Moroccan Darija (DODa) This project focuses on the identification and extraction of non-Arabic loanwords from the **Darija Open Dataset (DODa)**. Using an iterative approach combined with Unix-based automated extraction, we expanded a small seed list of borrowings into a broader lexicon of unique loanword tokens. ## Results Summary Through four rounds of iterative extraction, we processed 3,995 sentences and identified **487 unique loanword tokens**. ## Methodology The project follows the theoretical framework established by **Heath (1989)**, distinguishing between spontaneous code-switching and integrated borrowings. The extraction pipeline uses Unix stream processing: 1. **Initial Annotation:** Manual identification of loans in a subset of DODa. 2. **Expansion:** Using `grep` with whole-word matching to identify known loans in the remaining corpus. 3. **Deduplication:** Comparing vocabularies across rounds to isolate new lexical entries. ## References * Heath, J. (1989). *From Code-Switching to Borrowing: Foreign and Diglossic Mixing in Moroccan Arabic*. * DODa (Darija Open Dataset) Corpus. ## Author **Mohammed El khamlichi**