Analyzing and comparing on-line news sources via (Two-Layer) incremental clustering

In this paper, we analyse the contents of the web site of two Italian news agencies and of four of the most popular Italian newspapers, in order to answer questions such as what are the most relevant news, what is the average life of news, and how much different are different sites. To this aim, we have developed a web-based application which hourly collects the articles in the main column of the six web sites, implements an incremental clustering algorithm for grouping the articles into news, and finally allows the user to see the answer to the above questions. We have also designed and implemented a two-layer modification of the incremental clustering algorithm and executed some preliminary experimental evaluation of this modification: it turns out that the two-layer clustering is extremely efficient in terms of time performances, and it has quite good performances in terms of precision and recall.