
This dataset provides a trilingual lexicon extracted from Nam Phong Tạp Chí (南風雜誌), a Vietnamese literary and cultural journal published from 1917 to 1934 in Hanoi. Each issue included a lexicon supplement (Tự-Vựng, 字彙) defining new vocabulary in three scripts: quốc-ngữ (romanised Vietnamese), chữ nho (Chinese and Nôm characters), and French.
It covers issues 1 to 10, 13, and 14, representing 1,512 lexical entries across 12 CSV files. Transcription was performed using a Vision Language Model (Qwen VL 2.5) with fine-tuning, followed by manual correction. Transcription quality was evaluated using KAMI (CER: 11.48%, Word Accuracy: 87.94%, combining all options).
This dataset is produced as part of the RAG'it project, supported by DISTAM (DIgital STudies Africa, Asia, Middle East) and Calfa Vision. Full documentation is available in the README file included in this dataset.