Essay Quality Evaluation Using BART Transformer in English Writing Assessment Applications

Assessing the quality of essays and writing is a fundamental part of English writing assessment applications, easing the process for educators and automated systems to provide reliable feedback. Recent developments in natural language processing have fostered the growth of transformer-based models to study and improve the essay scoring and feedback process. Regardless, the existing methods of evaluating essays mean the recognition will have to overcome a number of challenges, including the subjectivity of the scoring process, a lack of view of possible writing styles, and a lack of understanding of contextual meaning, all of which restrict assessment reliability and fairness. Conventional automated scoring systems are typically found to have a lack of deep semantic understanding, which leads to lower levels of accuracy when evaluating. Recognizing and overcoming these challenges is the motivation for the evaluation of Essay Quality using Bidirectional and Auto-Regressive Transformers (EQ-BART) framework put forward in this study, where we made use of the BART transformer architecture in that it has a unique hybrid architecture, which includes bidirectional encoding and an auto-regressive decode, that can create a deep representation of an essay for the quality assessment of essays related to the essays's content, structure, and coherence. EQ-BART is able to maintain contextual information that is better for nuanced quality assessment beyond surface meaning (metrics). In addition, here evaluated the proposed EQ-BART framework using effective datasets on a scale, achieving greater qualirt analysis of 92%, essay score 5, quadratic weighted kappa analysis of 91.9%, essay quality, and human score vs predicted score analysis of 94.9% and assessing essays as compared to existing methods.

Ҳали таржима қилинмаган