Show GN: Garu: 브라우저에서 돌아가는 1.7MB 한국어 형태소 분석기 (F1 95.3%, WASM)

Garu brings morphological analysis to the browser using WebAssembly, solving a critical gap where existing tools (Kiwi, MeCab-ko) required server infrastructure. By combining code-based lookup with Viterbi decoding, it achieves 95.3% F1 accuracy while maintaining just 1.7MB model size—enabling offline, privacy-preserving Korean NLP directly in web applications. This represents a significant democratization of advanced language processing for client-side development.