Design and Development of Hybrid Learning Model for Intelligent English Writing Scoring System in an Educational Institution

In the modern education system, the evaluation mechanism have simple design and objective scoring, however with continuous scaling of educational content, the conventional manual scoring method face enormous pressure and certain subjective biases. To overcome this challenge, this research investigates an intelligent scoring mechanism for English writing at the institutional level using hybrid artificial intelligence algorithms to improve the accuracy and efficiency of evaluation. In this context, a novel hybrid deep learning framework integrating a Bidirectional Encoder Representations from Transformers (BERT) structure with a BiLSTM-Attention mechanism is introduced. This approach utilizes contextual embeddings, sequential relationships, and attention-weighted mechanism to evaluate grammar, coherence, lexical density, and argumentative structure. The proposed learning model was validated with Automated Student Assessment Prize Datasets and its performance has been compared with the existing models. Comparative experiments demonstrate that the proposed BERT-BiLSTM-Attention model for intelligent English writing evaluation significantly outperforms baseline models achieving an accuracy of 99.3% and an F1-score of 98.8%. The system proposed shows better performance in managing varied writing styles, minimizing human intervention, and offering instant scoring with clear feedback, thus representing a revolutionary leap in intelligent educational assessment.