Broadcast News Story Segmentation Using Conditional Random Fields and Multimodal Features

This paper proposes to integrate multi-modal features using conditional random fields (CRF) for broadcast news story segmentation. We study story boundary cues from lexical, audio and video modalities, where lexical features consist of lexical similarity, chain strength and overall cohesiveness, acoustic features involve pause duration, pitch, speaker change and audio event type, and visual features contain shot boundary, anchor face and news title caption. These features are extracted in a sequence of boundary candidate positions in the broadcast news. A linear-chain CRF is used to detect each candidate as a boundary/nonboundary tags based on the multi-modal features. Important inter-label relations and contextual feature information are effectively captured by CRF’s sequential learning framework. Story segmentation experiments show that the CRF approach outperforms other popular classifiers, including decision tree (DT), Bayesian network (BN), naive Bayesian classifier (NB), multi-layer perception (MLP), support vector machines (SVM) and maximum entropy (ME) classifier. key words: Story Segmentation, Conditional Random Fields