Skip to main content

Table 6 Evaluation result for each model

From: Finding epic moments in live content through deep learning on collective decisions

Type

Model

Accuracy

AUC

Train

Dev

Test

Train

Dev

Test

Feature-based

Chat frequency

0.6480

0.6555

0.6545

0.6963

0.7165

0.7159

Chat content

0.6857

0.6820

0.6841

0.7477

0.7437

0.7449

Video content

0.5653

0.6110

0.5478

0.6117

0.6114

0.6159

Deep learning

Chat frequency

0.8099

0.8089

0.8125

0.8720

0.8716

0.8748

Chat content

0.7271

0.7057

0.7013

0.7969

0.7710

0.7699

Video content

0.8210

0.8097

0.8140

0.9094

0.8879

0.8901

Multi-modal

MINT

0.8328

0.8317

0.8261

0.9054

0.9074

0.9006

(w/o importance learning)

0.8237

0.8256

0.8173

0.9018

0.8964

0.8911