Date

2016

Author

Kurfalı, Murathan

Metadata

Show full item record

Item Usage Stats

193
views

0
downloads

Cite This

In discourse parsing, the sense prediction of the Implicit discourse relations poses the most significant challenge. The thesis aims to develop a supervised system to predict the sense of implicit discourse relations in Turkish Discourse Bank (TDB). In order to accomplish that goal, the discourse level annotations obtained from TDB are used. TDB follows the PDTB-2’s sense hierarchy and for all experiments within the current study, only CLASS senses are considered. As the primary experiment, the classifiers are trained on merely implicit discourse relations based on the several linguistically informed features, such as polarity and tense information, to detect the possible sentence structures characteristic to each CLASS level sense. In the secondary experiment, the effect of Explicit discourse relations on the sense prediction of Implicit relations is investigated. The motivation behind this experiment is to provide insight regarding the differences and similarities of these two type of discourse relations which is another challenging topic in the discourse research. The results indicate that implicit discourse relations manifest significant differences in terms of their sentence structure depending on their sense. It is also revealed that using Explicit discourse relations alters the performance of the classification radically which suggests that these two type of the discourse relations are structurally different from each other. 

Subject Keywords

Turkish language., Discourse analysis., Supervised learning (Machine learning).

URI

http://etd.lib.metu.edu.tr/upload/12620328/index.pdf 
https://hdl.handle.net/11511/26057

Collections

Graduate School of Informatics, Thesis