Geniş Metin Koleksiyonlarından İteratif Bilgi Çıkarımı Iterative Information Extraction from Large Text Collections

Abstract

— There are various methods about information extraction from large texts. One of them is method of templates. We developed an automatic system that aims to produce pairs which have semantic relation between them using templates. We worked with morphological resolved and unresolved datasets. We obtained better templates from morphological resolved dataset. In our experiments, we observed that if too many templates were used for producing pairs, accuracy of produced pairs decreased. Also, we obtain better results for fixed and more reliable templates with using growing datasets