ScholarOne - Interactive contextualized and personalized math word
problem generation in authentic contexts using GPT and its influences on
geometry learning
Abstract
Automatic question generation (AQG) has been researched extensively for
educational purposes. This study proposed an automatic system for
generating contextualized and personalized mathematic word problems
(MWP) in authentic contexts using the Generative Pre-trained
Transformers (GPT). It comprises (1) authentic contextual information
acquisition through image recognition by TensorFlow and augmented
reality (AR) measurement by AR Core, (2) a personalized mechanism based
on instructional prompts to generate three difficulty levels for
learner’s different needs, and (3) MWP generation through GPT. A
quasi-experiment was conducted using 52 fifth-grade students to evaluate
the effectiveness of the proposed AQG on their geometry learning
performances. The learning behaviors were analyzed concerning authentic
context, mathematics, and reflective behavior aspects. The results
revealed that students who learned with the proposed AQG outperformed
students who learned with a decontextualized way on geometry learning
performances. Moreover, it was found that learning behavior related to
authentic context (i.e., problem context understanding and identifying
contextual information) and mathematics (i.e., applying math concepts
and the total number of the correct solution for medium-level MWP)
significantly improved geometry learning performances. Meanwhile,
learners showed positive perceptions toward the proposed AQG. Therefore,
our proposed AQG is useful to promote geometry problem-solving activity
in an authentic context.