Hotspots of Topics from Time Stamped Document : The Mapreduce Way

Hotspots of a word/topic are time periods with a burst of activities in a time stamped document set. Identifying and analyzing hot spots of topics has been an important area of research. Finding hot spots of topics requires processing of contents of documents which is often time consuming. In this thesis, we explore MapReduce style algorithms for computing hot spots of topics. MapReduce is a distributed parallel programming model and an associated implementation for processing and analyzing large datasets. User specifies a map function that processes a key/value pair to generate a set of intermediate key/value pairs, and a reduce function that merges all intermediate values associated with the same intermediate key. Many real world tasks are expressible in this model and this thesis explores the feasibility of implementing the hotspot algorithm using MapReduce. We design map and reduce functions appropriate for preprocessing of documents, and the hot spot computation. We implement the functions in Hadoop (a MapReduce framework for Apache Foundation) and conduct several experiments to assess the benefits of MapReduce style implementation versus simple sequential implementation.