Novelty Cloud mines topics from TV broadcasts

Researchers at University of Oulu's MediaTeam Oulu research group have developed a method that detects and summarizes novelty topics from closed captions transcripts in broadcast TV news.

Method is demonstrated in a web service that detects 2012 national news topics and visualizes them with word cloud views. Novelty Cloud organizes news topics in monthly and weekly views. Users can also view topical excerpts from the news content.

Novelty Cloud is based on a new data mining technique that computes novelty values for topics with the help of natural language processing and machine learning techniques. Topics are organized in a word cloud structure based on their novelty value. The technique improves over typical word clouds, which are based on word popularity or frequency information. Novelty detection can be applied to many program genres and it can be used to produce unique descriptions of individual TV programs.

Novelty Cloud is part of experimental TV search service platform Kuukkeli-TV that indexes programs from seven broadcast TV channels in Finland. Service users can create word queries into indexed TV content, find similar programs across TV channels and browse recently broadcasted programs by category. Service platform contains index of 150 000 broadcasted programs. It is utilized in research and development of semantic multimedia services to improve non-linear access of user-relevant digital video data.

Last updated: 26.2.2013