JOINT TOPIC MODEL FOR CROSS-MEDIA NEWS SUMMARIZATION

摘要

A method for providing a complementary summary of news information. The method comprises retrieving a first group of relevant text sentences about an event from a first content source, such as a news media stream, and retrieving a second group of relevant text messages about the same event from a second content source, such as a social media stream or microblogs. The method analyses the retrieved text information and calculates a measure of complementarity between the text information in the first and second groups and ranks the text sentences from the first group against the text messages from the second group according to the measured complementarity between these individual text sentences and messages. Then two types of complementary summaries are generated based on different granularities of complementarity: (1 ) considering summary-level complementarity, the two summaries are generated from two groups of text in such a way that they are complementary as a whole; (2) considering sentence-level complementarity, the two summaries are produced from two groups of text with strict one-on-one complementary correspondence inside the news sentences and text messages that constitute the respective summary.