Extractive based automatic text summarization in Turkish texts

In this thesis study, sentence grading methods related to the use of extractive text summarization method in Turkish texts were carried out as an experimental study, and new methods were proposed in addition to the methods found in the literature. The news texts obtained from the contents of the national news sites were used as a data set, a summary study of three different users was carried out in this data set, and these summary studies were later reduced to a single result. This part, defined as the control data set, was used to evaluate the performance with the summary to be created. In addition to the sentence grading methods available in the literature, major vowel harmony, minor vowel harmony, and a hybrid model in which these two rules are tested together were proposed. Evaluation of the three proposed methods was measured with ROUGE and BLEU metrics, and it was observed that it gave better results than the traditional methods. The summarization study was handled as a classification problem in addition to the evaluation with two metrics, and four different classification algorithms in the literature were used in this study. In the study, the use of only the proposed methods, the use of only the traditional methods and the hybrid method in which both methods are used together were evaluated in three different sections and it was observed that the proposed methods gave the best result in three algorithms and the hybrid method gave the best result in one algorithm.