The paper presents an integrated methodology for the continuous monitoring of traffic from Twitter, to support analysts in the early detection and categorization of critical situations. The main purpose of the developed technology is to monitor the information flow and track sensible situations or look for unexpected events. The main architecture relies on specific technologies covering language detection, text-mining, clustering and semantic labeling. The concept of 'mission profile' allows the analyst to specify target concepts and features related to the scenario of interest. Upon clustering and semantic labeling of running traffic, a novel refinement algorithm retrieves and aggregates relevant contents that would otherwise escape plain semantic analysis. Experimental results in different scenarios prove the method effectiveness in enhancing content retrieval and in greatly enhancing efficiency in contents presentation to the analyst.
|Titolo:||Content-Adaptive Analysis and Filtering of Microblogs Traffic for Event-Monitoring Applications|
|Data di pubblicazione:||2015|
|Appare nelle tipologie:||04.01 - Contributo in atti di convegno|