CLASSIFICATION OF HIGH SCHOOL ENTRANCE EXAMS AND TEACHER-MADE EXAM QUESTIONS ACCORDING TO THE REVISED BLOOM’S TAXONOMY WITH COMPUTATIONAL LINGUISTICS METHODS

Tez Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Yıldız Teknik Üniversitesi, Eğitim Fakültesi, Yabancı Diller Eğitimi Bölümü, Türkiye

Tez Danışmanı: Burcu Ünal

Tezin Onay Tarihi: 2024

Tezin Dili: İngilizce

Özet:

Bloom's taxonomy and its subsequent revision, the Revised Bloom's Taxonomy (RBT), are widely used in program development, assessment, and evaluation procedures because they represent lower and higher-order cognitive processes. They have been highly valued and applied in the field of education for many years. Revised Bloom's taxonomy is used to assess and improve the English language course program's success during curriculum development and assessments. This study aims to build data processing techniques and machine learning classification methods to classify the English questions of high school entrance exams and teacher–made written English exams for the 8th graders applied in Türkiye in the context of Revised Bloom’s Taxonomy. The study analyzed seven hundred ninety-two English exam questions of 8th graders labeled according to Revised Bloom’s Taxonomy (RBT). Computational Linguistics methods were used in this study. Random Forest models, text vectorization techniques; Bag of Words (BoW), Term Frequency-Inverse of Document Frequency (TF-IDF), and Part of Speech (POS) analysis were implemented to predict labels of questions. The findings revealed that predictive models developed via the Random Forest classifier and BoW vectorization technique perform reasonably well for predicting levels of questions such as “procedural knowledge and apply”, “factual knowledge” and “apply” categories with F1 scores of 0.91, 0.89, and 0.86 respectively. Accuracies for the predictions of these categories were 0.94, 0.90, and 0.91 respectively. However, due to insufficient data predictive models could not provide sufficient results for the “analyze” and “evaluate” categories of cognitive process dimension and the “metacognitive knowledge” category of knowledge dimension. The findings in this study can be a benchmark for further studies regarding the classification of exam questions considering RBT.

Keywords: Revised Bloom’s Taxonomy, Computational Linguistics, English Questions, Corpus Linguistics, Machine Learning