Building News Timelines: from short term to long term memory

The increasing amount of online published news make it difficult for the users to cope with the news space. In large organisation, where the news can be an important input for decision making, the urge of intelligent systems organising and presenting the news is more and more present. Several approaches exist for news analysis, the most popular is the news aggregators, such as Google News, that present the last published news grouped into stories and categories. The main limitation of these news aggregation and analysis systems is that they are only taking into account the last published news, making sometimes difficult to contextualise a news story with past ones.
The system we are presenting aims at tackling this issue, by providing the users with an automatically built news timeline, that relate and put in context the current news stories with regard to the past ones. The system try to organise the news space as a memory system where the recent publications are managed is episodes in the short-term memory while the older ones are abstracted in long-term memory. The key aspect of the system is to keep the relevant links between the two mechanisms of storing the news.