Individualized Storyline-based News Topic Retrospection

It takes a great effort for common news readers to track events promptly, and not to mention that they can retrospect them precisely after it occurred for a long time period. Although topic detection and tracking techniques have been developed to promptly identify and keep track of similar events in a topic and monitor their progress, the cognitive load remains for a reader to digest these reports. A storyline-based summarization may facilitate readers to recall occurred events in a topic by extracting informative sentences of news reports to compose a concise summary with essential episodes. This paper proposes SToRe (Story-line based Topic Retrospection), that identifies events from news reports and composes a storyline summary to portray the event evolution in a topic. It consists of three main functions: event identification, main storyline construction and storyline-based summarization. The main storyline guides the extraction of representative sentences from news articles to summarize occurred events. This study demonstrates that different topic term sets result in different storylines, and in turn, different summaries. This adaptation is useful for users to review occurred news topics in different storylines