AUTOMATIC QUESTION GENERATION FOR 5W-1H OPEN DOMAIN OF INDONESIAN QUESTIONS BY USING SYNTACTICAL TEMPLATE-BASED FEATURES FROM ACADEMIC TEXTBOOKS

摘要

The measuring of education quality in school can be conducted by delivering the examination to the students. Composing questions in the examination process to measure students? achievement in the school teaching and learning process can be difficult and time consuming. To solve this problem, this research proposes Automatic Question Generation (AQG) method to generate Open Domain Indonesian Question by using syntactical approach. Open Domain questions are questions covering many domains of knowledge. The challenge of generating the questions is how to identify the types of declarative sentences that are potential to be transformed into questions and how to develop the method for generating question automatically. In realizing the method, this research incorporates four stages, namely: the identification of declarative sentence for 8 coarse-class and 19 fine-class sentences, the classification of features for coarse-class sentence and the classification rules for fine-class sentence, the identification of question patterns, and the extraction of sentence?s components as well as the rule generation of questions. The coarse-class classification was carried out based on a machine learning with syntactical features of the sentence, namely: Part of Speech (POS) Tag, the presence of punctuation, the availability of specific verbs, sequence of words, etc. The fine-class classification was carried out based on a set of rules. According to the implementation and experiment, the findings show that the accuracy of coarse-class classification reaches 83.26% by using the SMO classifier and the accuracy of proposed fine-class classification reaches 92%. The generated questions are categorized into three types, namely: TRUE, UNDERSTANDABLE, and FALSE. The accuracy of generated TRUE and UNDERSTANDABLE questions reaches 88.66%. Thus, the obtained results show that the proposed method is prospective to implement in the real situation.