Automated Rhetorical Device Detection in Student Essays for Argumentation Skill Training

Rhetorical devices Automated recognition of them in student essays is imperative in improving argumentation skills and in aiding efficient teaching of writing. Detecting such patterns as analogies, metaphors, and repetitions may give some information about the persuasion methods used by students and their general proficiency in arguments. Current systems to detect rhetorical devices may either be based only on rule-based systems or neural models, which may be limited to low adaptability to different styles of writing and lack of contextual information about subtle language. To overcome such problems, we suggest a Hybrid Rule-Based and Neural Sequence Labeling (HR-NSL) model. This method uses the accuracy of rule-based patterns of known rhetorical structures along with the contextual adjustability of neural sequence labeling models to identify implicit or compound devices. A combination of the two strategies will enhance the accuracy in detection and interpretability to facilitate learning by HR-NSL. The suggested approach may be implemented in automated feedback systems, where the teacher and learners can detect rhetorical devices in the essay, trace the progression of the argumentation process, and design specific interventions on the writing process. Experimental analysis proves that HR-NSL is more successful than the traditional rule-based and neural methods providing larger precision and recall of multiple rhetorical devices on various student essays. The findings suggest that such a mixed method will be useful in motivating argumentation skills in the learning context.

Ҳали таржима қилинмаган