Big Data-Driven Platform for Cross-Media Monitoring

Liana Napalkova, Pablo Aragón, Juan Carlos Castro Robles

Abstract: The abundance of online media content requires highly scalable architectures to allow cross-media monitoring. This paper presents an innovative big data-as-a-service platform for analysing large complex networks in order to enhance cross-media monitoring. In contrast to the existing media monitoring systems, the platform equips marketers with several distinctive features. First, while most of the systems perform quantitative exploratory analysis of social media, our platform applies graph analytics in order to reveal social interaction types, hidden patterns in the cross-media network and the information diffusion over time. Second, our platform integrates and implements distributed versions of graph analytics algorithms (Louvain, HITS and others) that can scale to a large volume of data. Third, the creation of cross-media graphs is triggered by user-defined queries that can be easily specified by marketers. Thus, end-users can build and analyse different graphs according to specific goals of the study. Finally, the platform allows reducing Hadoop cluster usage costs due to executing the graph mining algorithms on demand triggered by user-defined queries. Instead of running costly streaming processes that continuously listen for new queries, we implemented Spark-as-a-service approach via Apache Livy REST interface.