| Popis: |
In this study, we address the multi-document summarization challenge. We proposed a summarizer application that implements three well-known multi-document summarization techniques; Topic-word summarizer, LexPageRank summarizer and Centroid summarizer. The contribution in this study is demonstrated by proposing a fourth summarization technique that is built on the previous acquired knowledge and experiments performed on the previously mentioned summarization techniques. Evaluating the system-generated summaries is performed using ROUGE [1], results showed that the new summarizer outperforms the other summarization techniques, and it takes a relatively short time to generate summaries comparing to other summarizers. However, LexPageRank summarizer evaluation performed better than the new summarizer evaluation, the cost of achieving a better evaluation using this technique was the time needed to generate the summaries, LexPageRank summarizer needs a long time to generate summaries comparing to other summarizers. In this study, DUC04 is used as a corpus in testing and implementing the proposed application. |
