Parameters Affecting the Judgment Task While Summarizing Documents

Due to the instability in user agreement, necessity for improvements in document summary generation has paid huge attention. This paper, presents improvements to document summarization by analyzing the sentence position and recommendations of sentences from other sentences for making an efficient document summarizer. As we know humans are good in summarizing the contents, we have made few studies to improve the efficiency of our system much closer to user selection (manual summaries). This article focuses only on providing improvements for news articles. We have attempted to obtain summaries much closer or equal to manually generated summaries and the results obtained were promising. We also, show that term frequency approach combined with position weight gives better results, while adding node weight with the above feature yield results that were significantly better than the former approach. The paper also, illustrates some studies on some common evaluation criterion to generate a unique summary by various users. The results were also, compared with commercially existing Microsoft summarizer. The results produced by us were better as compared to the existing summarizers