Propagation in bipartite graphs for Topic Extraction in Data Streams

Abstract

A lot of of research is devoted to dynamic data mining, which generally contains large amounts of solid, multi-dimensional and high speed datas. Such data are common in everyday applications, such as telephone calls, retail sales, performance, data center, manufacturing operations, among others. A instance particular problem is exploration and mining of textual data stream collections. This relevant area, which analysts are interested in behavior perception and data trends, still imposes many challenges to researchers. In this context, this project aims to investigate innovative text mining techniques for topics extraction from data stream. Large collections of text data are considered, typically news contents, Web-logs etc. The publications of these data happen in real time, which results in constant topics changes distribution. The scope of this proposal includes all the text mining steps related to topics extraction from stream and, considering concepts drifts. To do this, will be explored and developed text mining technologies for extraction and identification of dynamic topics, information extraction over time and texts categorization for temporal coverage. The results of this PhD project will be validated through benchmark and techniques available for text mining and Data Streaming.