Entropy and optimization based extractive text summarization in text graphs

The increase in text sizes requires the increasing the need for automatic document summarizing tools. In the thesis, two new methodologies for summarizing text documents; Karcı Summarization, Multi-Objective Sapling Growing Up Algorithm (MO-SGA) are offered. Karcı Entropy was used in a document summarizing study with a unique approach and it was introduced to the literature. The proposed system does not require training data. Optimization techniques are another technique commonly used in summarizing studies. However, in these studies, the different contradictory objectives are often weighted or combined into a single-purpose problem. This conversion directly affects the quality of the solution. In addition to Karcı Summarization, the thesis presents an approach in which the two main characteristics of a rich summary, maximum coverage and minimum redundancy, are modeled as the objectives of optimization. With the added pruning step, the Sapling Growing Up Algorithm was moved one step further and MO-FGA was introduced. A simultaneously optimization is implemented within the scope of the presented approach. The results of the proposed approaches were tested with open access data sets (DUC 2002-2004). The performances of the approaches were examined using Recall-Oriented Understudy for Gisting Evaluation (ROUGE) criteria. Experimental results reveal that there are promising techniques in both approaches.