Achieving More Coherent Summaries in Automatic Text Summarization; an Ontology-based Approach

Abstract

Since the growth of information available on the internet has grown out of hand, automation of text summarization process has become more important. Summaries that are in the form of a condensed version of original text, containing its important information, and considered as a good alternative to reading the original text. One of the main requirements of the machine produced texts, is their coherence and the semantic relation between their sentences. Reading a non-coherent summary, in spite of that does not help the readers become aware of the information in the original text, it also creates confusions in their mind. The main purpose of this paper is how to achieve more coherent summaries in automatic text summarization process. For this purpose there has been a system designed, that using the concepts of ontology, automatically summarizes Persian documents, and tries to produce more coherent summaries. In this light, a technique has been devised that based on it, presence of a sentence in the summary, increases the probability of choosing its adjacent sentences. The FarsNet ontology is the basis for ontology-based calculations in this paper. The results show that the suggested approach succeeds in producing coherent summaries.