Papers by Shengli Hu

Copied to clipboard

Challenge:	sommeliers have three skills: wine theory, blind tasting, and beverage service . current study suggests that the sophist profession is at least to some extent automatable .
Approach:	They propose to train machine learning models that match sommelier's skills and compare results with real data.
Outcome:	The proposed models outperform human sommeliers on most tasks, compared with real data from a large group of wine professionals.

Copied to clipboard

Challenge:	despite the importance and potential impact of detecting concealed information, research on detecting it has been scarce.
Approach:	They propose a multi-task learning framework that automatically detects concealed information from text and speech using acoustic-prosodic, linguistic, and individual feature sets.
Outcome:	The proposed framework outperforms human performance by 15% in acoustic, linguistic, and individual features.