Since the birth of web 2.0, users no longer just consume, but are now active creators of content that is going to be consumed by other users. This new dynamic took data generation to a whole new scale, called planetary scale or web scale. Often, this data represents relationships between its elements, such as in social networks, recommendation systems, online boards, email networks, scientific citation networks and others. Therefore, this data can be properly modeled as graphs, which contain properties like weights on its edges, associated texts and dynamic behavior that can be explored. To analyze such graphs, the main approach consists of using distributed processing techniques via computer clusters leading to high costs and techinical complexity that can be prohibitive. So, it is desirable to be able to process planetary scale graphs using only a single computer. To do this, we intend to combine edge and vertex centric iterative processing with discrete matrix processing and text processing techniques aiming to develop an analysis framework capable of recognizing patterns, comprehension and helping with decision making. With such methods we intend to develop new algorithms and systems to solve problems like fraud detection, behavior analysis and sentiment analysis, in a variety of domains.
News published in Agência FAPESP Newsletter about the scholarship: