An enhanced spectral based approach for initial and update summary generation

Update summarization is a recent trend in summarization task. The main goal of update summarization is to generate an update summary which includes the sentences from a collection of documents representing the evolved information. It is necessary that the selected sentences must be different from the previously selected sentences and also expected that some similarity in information be present. Previous update summarization approaches utilizes different clustering and ranking techniques to produce an update summary. However, these clustering and ranking techniques are considered as a separate task. This paper introduces an enhanced spectral technique for generating an initial and update summary. The spectral information has been taken to integrate clustering and ranking process. To evaluate the performance of proposed framework, machine generated summaries can be compared against human summary.