Softverski sistem za određivanje semantičke sličnosti kratkih tekstova na srpskom jeziku


This paper describes a software system for determining the degree of semantic similarity of two short texts written in Serbian. Its basic working principles are presented, as well as the phases of its functioning and evaluation. It also describes the process of generating a paraphrase corpus, which was used for evaluation purposes. Finally, evaluation results are discussed and further improvements of the system’s precision are considered.

Proceedings of the 19th Telecommunications Forum (TELFOR 2011), Belgrade, Serbia, pp. 1249-1252, IEEE