linuxscout/algerian-name-transcription-corpus

Aldjia: Algerian-name-transcription-corpus # Algerian Name Transcription Corpus The **Algerian Name Transcription Corpus (Aldjia علجية)** is an open bilingual corpus of Algerian personal names. It provides aligned Arabic and French spellings of first names and family names collected from real administrative data. The corpus is intended for research and applications involving Arabic–Latin transliteration, named entity recognition, information retrieval, record linkage, and multilingual natural language processing. ## Contents The corpus is distributed as two CSV files. ### First Names `first_names.csv` | Column | Description | | -------------- | ------------------------------------- | | `firstname_fr` | French (Latin) transcription | | `firstname_ar` | Arabic spelling | | `sexe` | `feminin` or `masculin` | | `generation` | `gen1` (parents) or `gen2` (children) | Example: ```text firstname_fr,firstname_ar,sexe,generation Mohamed,محمد,homme,gen2 Fatima,فاطمة,fem,gen1 ``` ### Family Names `family_names.csv` | Column | Description | | --------------- | ---------------------------- | | `familyname_fr` | French (Latin) transcription | | `familyname_ar` | Arabic spelling | Example: ```text familyname_fr,familyname_ar Benali,بن علي Amrani,عمراني ``` ## Corpus Generation The corpus was generated from Students registry data. The generation process includes: 1. Extraction of relevant columns from Excel files. 2. Construction of bilingual name pairs. 3. Removal of duplicate entries. 4. Data normalization. 5. Automatic quality verification. 6. Generation of corpus statistics. ## Validation Several automatic checks are performed to detect potential errors, including: - inconsistent spacing between Arabic and French names; - Latin letters appearing in Arabic fields; - Arabic letters appearing in French fields; - duplicate entries; - additional validation rules can easily be ad …