Topic clustering and topic evolution based on temporal parameters

Historical documents contains extreme amount of information about past events, often in unstructured form. Once dates and document names are identified that can differ by genre, we examine collections to detect events. Temporal text mining (TTM) is used to ascertain the temporal patterns in text information collected over time. Trend analysis from the stream of text documents generally uses an approach based on topic detection and tracking (TDT). The task of topic detection is used to detect topics that are previously unknown to the system. Tracking generates the evolution of each topic over the period of arrival time. In this work the TDT task has been formulated as a clustering problem in a class of self-organizing neural networks, called the Adaptive Resonance Theory (ART) networks. We also propose that our algorithm has been able to detect hot topics automatically and track them with good accuracy. From our experimental studies we prove this by comparing the effectiveness of the different validity indices of simple k-means clustering method. We also show the benchmarking results of different kinds of datasets.