Affiliations: [a] School of Computer Science, Wuhan University, Wuhan, China
| [b] Department of Computer Science, Binghamton University, Binghamton, USA
Correspondence:
[*]
Corresponding author. Wei Yu, School of Computer Science, Wuhan University, 430072 HB, Wuhan, China. E-mail: [email protected].
Abstract: With the daily release of the huge amount of new information, the web information such as video, image, audio and text information is growing dramatically. Most of this vast information changes over time. It may become ineffective, obsolete and worthless, and even affect user’s understanding of web information, degrade their experience. Therefore, analyzing this temporal-sensitive information in multimedia is a vital issue. To this end, this paper analyses the temporal sensitivity of multimedia web information to find what kind of information is temporal-sensitive, how this information is responsive to time, and how to evaluate the temporal sensitivity of information. We start with four types of features, that is time, content, user behavior and related multimedia news, then set up a triple model to depict news. By establishing the energy transfer relationship between news and related and similar news, time, and user behavior, we measure the energy, and use the change ratio of energy as the temporal sensitivity of news. The data set is a multimedia news corpus, including video, image and text news. In the experiment, we take the users’ comments as the validation set. The result basically matches the validation set, and it shows our metric is reasonable.
Keywords: Big data analysis, temporal metric, temporal sensitivity, multimedia news
