Metadata Audio Data Retrieval TRIZ Cases
- Voice Query Pronunciation Prediction for Ambiguous Entity Search
Selective pronunciation prediction preserves ambiguous terms during voice-to-text conversion, improving entity search accuracy.
- Video Music Matching Through Image-Based Soundtrack Recommendation
Image recognition and video attributes guide soundtrack recommendations, improving match accuracy and reducing manual search during editing.
- Multi-Modal Feature Recognition for Reliable Audio-Video Tagging
Independent video, audio, and text encoders combine through a neural mixer to improve feature classification accuracy and reduce manual tagging.