[Abstract]

Coherent Arrangement of Sentences Extracted from Multiple Newspaper Articles

Naoki Okazaki (The University of Tokyo, Japan),
Yutaka Matsuo (Cyber Assist Research Center AIST Tokyo Waterfront, Japan, Japan)
and Mitsuru Ishizuka (The University of Tokyo, Japan)



Multi-document summarization is a challenge to infor-mation overload problem to provide a condensed text for a number of documents. Most multi-document summarization systems make use of extraction techniques (e.g., important sentence extraction) and compile a summary from the se-lected information. However, sentences gathered from mul-tiple sources are not organized as a comprehensible text. Therefore, it is important to consider sentence ordering of extracted sentences in order to reconstruct discourse struc-ture in a summary. We propose a novel method to plan a coherent arrangement of sentences extracted from multiple newspaper articles. Results of our experiment show that sentence reordering has a discernible effect on summary readability. The results also shows signi.cant improvement on sentence arrangement compared to former methods.