Innovative Multi-View Approach for Tracking Topics in News Videos

[ 31 May 2021 | vol. 14 | no. 1 | pp. 11-22 ]

About Authors:

Majida Meshari
-Tikrit University, Iraq

Abstract:

Existing research on tracking topics in news videos typically requires a large number of labeled examples. However, the process of labeling videos is time-consuming, making it difficult to generate sufficient labeled data for practical applications. In this paper, we propose a novel approach to track topics in news videos using only a few labeled samples. The main features of our proposed approach are as follows: (1) We employ a multi-view learning process to address the issue of insufficient labeled training samples during topic tracking. (2) We utilize Principal Component Analysis to simplify the feature vectors and enhance processing speed. (3) We introduce a collaborative training process that allows tracking of related news videos using just one labeled sample. Experiments conducted on our collection of Chinese news videos demonstrate that our approach achieves commendable results, with a precision of 82.5% and a recall of 96.4%.

Keywords:

Topic tracking; Multi-view learning; Principal Component Analysis; Collaborative training

 

About this Article: