Knowledge Base Driven Automatic Text Summarization using Multi-objective Optimization

Abstract

Automatic Text summarization aims to automatically generate condensed summary from a large set of documents on the same topic. We formulate text summarization task as a multi-objective optimization problem by defining information coverage and diversity as two conflicting objective functions. With this formulation, we propose a novel technique to improve the performance using a knowledge base. The main rationale of the approach is to extract important text features of the original text by detecting important entities in a knowledge base. Next, an improvement on the multi-objective optimization algorithm is also proposed for the automatic text summarization problem. The focus is on improving efficiency of the each steps in the evolutionary multi-objective optimization process which is applicable to all tasks with the same problem formulation. The result summary of the suggested method ensure the maximum coverage of the original documents and the diversity of the sentences in the summary among each other. The experiments on DUC2002 and DUC2004 multi-document summarization task dataset shows that the proposed model is effective compared to other methods.

Text Summarizing and Clustering Using Data Mining Technique

“…A multidocument extractive process can be precisely formulated as extracting significant text units from multiple related documents, eliminating redundancy, and rearranging the units to produce an efficient summary. An alternative approach to ensure good coverage and avoid redundancy is a clustering-based approach that groups similar text 59 units (paragraphs, sentences) into multiple groups to identify common information themes and selects text units one by one from clustering to the final summary [6]. Each block consists of similar text units representing a subtopic (theme).…”