Utilizing external information on contents for broadcasting services

This paper proposes a novel method to construct the content knowledge from blogs. Knowledge graph is constructed with images and keywords extracted from blogs and became the content knowledge by connecting images in the graph to frames in the content. In connecting process, distance with timeline locality matches images and frames based on singular feature for scenes. Then, matching results are re-ranked with ORB to reflect other information. In experiments with Korean TV series, the proposed method achieves 73.7% of accuracy in matching. The content knowledge is applied in car identification and improves the performance of baseline up to 73%.