Cross-lingual retrieval of identical news by using image and text information

Recently, the importance of archiving news videos has increased. Therefore, for reusing large quantities of accumulated news videos efficiently and effectively, technology for news topic retrival and tracking is necessary. Moreover, in order to understand a certain topic from various viewpoints, the opportunity for viewing news programs from all over the world has increased. Therefore, we focus on identical event detection in various news programs from different countries. However, cross-lingual retrieval using text information is difficult due to the performance of machine translation and moreover, different viewpoints and cultures. In this paper, we propose a cross-lingual retrieval method for detecting identical news topics that exploits image information together with text information. We utilize the existence of near-duplicate video segments as a strong information for the purpose. In an experiment, we verified the effectiveness of making use of the existence of near-duplicate video segments and the possibility of improving retrieval performance by using them.