Weighted Archetypal Analysis used for Text Summarization

Abstract: This paper demonstrate a summarization system that generates a summary for a given documents based on sentence similarity measures using weighted archetypal analysis. Most of the former approaches for multi document summarization give the summary by using only the Query Focused methods applied on the given document. And many approaches like matrix factorization methods to search either low rank approximation method or hard/soft clustering methods to get better document summary for the given documents. In this paper we propose different method called Weighted Archetypal Analysis an efficient approach for multi document summarization using extractive type summarization which uses the term frequency for sentence importance measures: Frequency of the terms i.e. archetypes, values in the sentence. The sentences are ranked according to their respective weights (scores) and the top rank (highest weight) sentences are selected for summary. The summary is generated by using Weighted Archetypal Analysis to compute archetypes, term frequency and significant sentences to the target documents evaluation measure.

Keywords: Generic document summarization, weighted archetypal analysis, Text summarization, Matrix factorization approach, Term Frequency.