Hi PH! I'm Kunal — a developer from Uttarakhand, India. I built TranscriptAI because Japanese business communication almost never says "no" directly. 「検討させてください」literally means "let me consider it." It means no. Most NLP systems get this completely wrong. I spent 2 months studying Japanese corporate linguistics and built 16 nemawashi detection patterns with confidence scoring, keigo register detection via MeCab morphological analysis, and cross-script speaker identity resolution (田中 ↔ Tanaka ↔ Director). Also handles Hindi indirect speech — 8 patterns, Devanagari and Roman both. Accuracy went from 22% → 93% across 5 iterations. Happy to answer any questions — especially on the Japanese NLP side. That's where most of the hard work happened. 🔗 Live demo: huggingface.co/spaces/KunalTheBeast/TranscriptAI 💻 Code: github.com/aiKunalBisht/Transcript-ai

TranscriptAI

Meeting AI that understands Japanese soft rejections

Meeting AI that understands Japanese soft rejections