<![CDATA[Test of English as a Foreign Language (TOEFL) is one of the evaluations requiring good quality of the questions so that they can reflect the English abilities of the test takers. However, it cannot be denied that making such questions with good quality is time consuming. In fact, the use of computer technology is able to reduce the time spent in making such questions. This study, therefore, develops a model to generate error identification typed questions automatically from news articles. Questions from the sentences on news sites are created by utilizing Natural Language Processing, Levenshtein Distance, and Heuristics. This model consists of several stages: (1) data collection; (2) preprocessing; (3) part of speech (POS) tagging; (4) POS similarity; (5) choosing question candidates based on ranking; (6) determining underline and heuristics; (7) determining a distractor. Testing ten different news articles from various websites, the system has produced some error identification typed questions. The main contributions of this study are that (i) it can be used as an alternative tool for generating error identification typed questions on TOEFL from news articles; (ii) it can generate many questions easily and automatically; and (iii) the question quality are maintained as historical questions of TOEFL.]]>
Copyrights © 2020
