ENG
INA
84th Doctoral Promotion
Promovendus I Made Widiartha, S.Si., M.Kom passed his doctoral dissertation and became the 84th PhD in the Study Program of Doctoral Engineering Science (PSDIT). The dissertation titled Multi-Document Summarization Model Using Markov Clustering, Tuna Swarm Optimization, and Similarity Based Ordering received a very satisfactory predicate.
Dr. I Made Widiartha, S.Si., M.Kom describes that the need for automated document summarization technology is currently increasing. This is because the existence of websites as a medium for searching and spreading information causes rapid increase in text data. People generally use search engines to search for information on the Internet. The result achieved is a list of website links that the user must open all other links to find the information the user contains. In many cases, the content displayed on various websites, or even the same information between one website and another, or even the majority of information obtained, has nothing to do with what the users want. A solution is needed to overcome this problem so that users can receive relevant accurate information from their website pages.
Although various approaches have been used to create multi-document summary models, issues of accuracy, redundancy and coherence remain the focus of research questions. In this study, a new multi-document summarization model has been built by utilizing the collaboration of several methods, namely Tuna Swarm Optimization (TSO) for sentence weighting, two-stage markov clustering (MCL) for handling redundant sentences, and similarity-based ordering for sentence ordering. The developed model is based on supervised learning which utilizes datasets for the training and testing stages of model performance. The dataset amounted to 500 documents collected using the scraping method on ten Indonesian language online media. This dataset is divided into two parts, namely 80% as training data and 20% as test data.
The study results showed that the proposed model performed better in terms of ROUGE value than several approaches, such as the BAT Algorithm, Whale Optimization Algorithm, and Artificial Bee Colony. This performance was obtained by optimizing the TSO parameter values. The optimal value for parameter a = 0.7, z = 0.9, and the optimal number of tuna > 80. Two-stage Markov clustering model is effective in handling redundancy. This can be seen from the summary results which have 8.1% better performance than without using Markov clustering. In terms of coherence level, the developed similarity based ordering model has better performance compared to the other two methods, which is 23.6% higher than Weighted Based Ordering and 32.1% higher than Chronological Ordering. The performance of this model has been validated using K Fold Cross validation and the results show the consistency of the model's reliability in various multi-document characteristics. The promoter, co-promoters I and II, are Prof. Ir. Rukmi Sari Hartati, MT., Ph.D., Dr. Ir. Dewa Made Wiharta, ST., MT., and Dr. Nyoman Putra Sastra, ST., MT, respectively.
FACULTY OF ENGINEERING