Sentence Ordering Algorithm with Subject Criterion for Automatic Multi-Document Summarization

In multi-document summarization, order of sentences in summarization result must be coherence and it must represent information in correct steps to make it easy to understand by the reader. Problem arises when some subject of sentences are represented using pronouns. The subject pronoun in an incorrect sentence order will confuse the reader as the pronoun can refer to more than one subject. In this paper, we propose a new subject criterion for sentence ordering strategy to complement the existing ordering strategy. Sentences will be clustered based on its subject and will be ordered with respect to subject levels. We test the system using Document Understanding Conference summarization data and compare it with results from existing algorithm without using subject criterion. The accuracy of ordering with all criterions including subject criterion is 83%. The result shows that there is a slight improvement in ordering accuracy when subject criterion is included.