Abstract: Nowadays many online social networking services that enable the user to send and receive the information. These information are being shared at an extraordinary rate and their unrefined form, although providing useful information and also be vast. It is difficult for the end users and data analyst to rectify huge amount of noise and data redundancy which included in millions of text. To ease the problem, novel continuous single and multi topic summarization framework has been proposed for text streams. Traditional summarization systems mainly focus on static and small-sized data sets, so, there are not efficient as well as scalable for huge amount of data sets and data streams. Their iterative/recursive results are insensitive to time and difficult to detect topic evolution. Our proposed framework is efficiently designed to deal with dynamic, fast arriving, and large-scale text streams and multi topic summarization. Our framework consists of clustering, single and multi topic summarization and evolution techniques to generate text. A novel clustering algorithm has been proposed to cluster texts and maintain distilled statistics in a data structure. Next a single and multi topic summarization technique has been proposed for generating online summaries and historical summaries of indiscriminate time durations. By comparing manually created summaries and summaries created by some important traditional summarization systems to evaluate the generated summaries efficiently. And finally, an effective topic evolution detection method has been proposed which automatically produce the timelines by monitoring different variations from text streams.

Keywords: Clustering, continuous summarization, single and multi topic summarization, summary, timeline generation, Text stream