ViewPulseAI/viewpulse-vpa-sample
收藏资源简介:
View Pulse AI (VPA) 多模态媒体行为数据集是一个专有数据集,专为预测观众行为分析而设计。它连接原始视频内容与实时观众参与数据,优化用于AI驱动的评分预测引擎。数据集采用Parquet格式,包含音频紧张度与氛围频谱特征、上下文情感分析的NLP向量、场景构图与动态的视觉特征,以及基于序列的时序行为模式特征。标签为Reaction_Class(-1, 0, 1),用于观众行为分类。该数据集通过SBV(场景行为向量)引擎高精度预测观众保留率,支持数据驱动的内容制作和战略广播。
The View Pulse AI (VPA) Multimodal Media Behavior Dataset is a proprietary multimodal dataset designed for predictive audience behavior analysis. It bridges raw video content with real-time viewer engagement, optimized for AI-driven rating prediction engines. The dataset is in Parquet format and includes features such as audio tension and atmosphere spectral features, contextual sentiment analysis via NLP vectors, scene composition and dynamics in vision features, and sequence-based temporal behavioral patterns. The label is Reaction_Class (-1, 0, 1) for audience behavior categorization. It is utilized by the SBV (Scene Behavior Vector) engine to predict viewer retention with high precision, enabling data-backed content production and strategic broadcasting.




