부산대학교 도서관

Natural Language Processing (NLP) offers methods for understanding and quantifying the similarity between written documents. Within the testing industry these methods have been used for automatic item generation, automated scoring of text and speech, modeling item characteristics, automatic question answering, machine translation, and automated referencing. This paper presents research into the use of NLP for the identification of enemy and duplicate items to improve the maintenance of test item banks. Similar pairs of items can be identified using NLP, limiting the number of items content experts must review to identify enemy and duplicat items. Results from multiple testing programs show that previousely unidentified enemy pairs can be discovered with this method. [ABSTRACT FROM AUTHOR]