A Method to Extract Sentences Referenced by Students' Technical Reports Using Parse Trees and Word Concepts

We are developing a system that automatically assesses Japanese technical reports submitted by university students. An important function of the system is to evaluate the accuracy of references in the reports. In this paper we propose a method to extract the sentences referenced in a report from the actual reference. The method extracts sentences using parse trees and word concepts from the reference. The experimental result shows that the fourth cumulative extraction rate of the method is about 90%.