Applying ML at scale for New User Experiences
summary
Describes how Samsung’s Bixby Voice uses ASR and NLU to enable voice interactions across Samsung and third‑party services, covering R&D work including India labs. Presents challenges and deep‑learning approaches for domain/intent classification (CNN vs RNN, word vs character models, handling data variation/conflicts) and two ASR improvement techniques: a hierarchical accent‑determination system with accent‑specific models (e.g., GB, US, Australian, Canadian, Spanish, Korean, Indian, Chinese) and on‑device user‑specific acoustic model adaptation including utterance selection and noisy ground‑truth handling.
AI-written summary · about
related talks
39mIt's All in the Data: The ML Behind Alexa's AI Systems
47mBuilding smarter apps with Machine Learning, from magic to reality
47mBuilding a cloud AutoML platform for creating voice robots
32mHarnessing Large Language Models for Enhanced Customer Experience
36mVoice Enabling Mobile Apps and Custom Hardware Devices
08mQ n A with Dr Ananth Sankar
31mPower-Up Your App With Apple’s Machine Learning Frameworks
14mAI for a Billion Voices: Taking AI where there is Maximum Impact