‘The Voice’: New Challenges in a Zero UI world
summary
Pandora launches an in-app voice mode serving part of a 100+ million monthly user base and classifies queries as known-item, thematic, or broad open-ended. It uses deep BiLSTM+CRF slot-filling for intent and tags, phonetic and fuzzy retrieval plus phonetic search with personalization via embeddings to rank a single play result, asks for confirmation on low-confidence matches, and runs a Bernoulli beta bandit using playtime reward to balance exploration under strong single-item selection bias.
AI-written summary · about
chapters (15)
- 0:00Start
- 0:23Pandora
- 1:00Voice Mode
- 2:00Search and Recommendation
- 3:21Known-item Queries
- 4:37Search Challenges
- 6:49Personalized Entity Search
- 8:38UX Considerations
- 9:28Thematic Queries
- 10:22Thematic Recommendations
- 12:50Natural Language Understanding
- 13:41NLU Architecture
- 14:46Broad Queries
- 15:20Bandit Solution
- 17:24Other Considerations
related talks
27mPlay
12mVOICE Summit 2019 | VED Talks | John Gillilan | July 23
35mHaving a Voice on Android: How hard can it be?
17mReimagining Customer Experiences Integrating Voice into Digital Strategy
59mDesigning for speech
50mSpeak To Me: Voice App Conversation Practices
44mGame Changing Voice Assistant: Industry Perspectives
07mVOICE 2021 | Elise Pinto, Vocads