APPLICATION OF NLP ALGORITHMS IN AUTOMATED ASSESSMENT OF ENGLISH WRITING

Abstract

Automated assessment of English writing has gained increasing significance in educational settings due to the growing need for scalable, consistent, and objective evaluation mechanisms. This study investigates the application of Natural Language Processing (NLP) algorithms in automated writing evaluation (AWE), tracing the historical evolution from early surface-feature approaches to modern transformer-based systems that leverage deep contextual embeddings, syntactic parsing, semantic analysis, and discourse modeling. Drawing on a comprehensive review of both foundational systems (e.g., PEG, e-rater) and recent neural architectures, we analyze how these tools assess grammar, vocabulary richness, coherence, cohesion, semantic relevance, and organizational quality. The discussion also highlights key methodological considerations, such as tokenization, parsing, semantic role labeling, coherence modeling, and feedback generation, and reflects on the ethical and pedagogical challenges — including bias, overemphasis on formulaic writing, and fairness in EFL contexts. Finally, the paper explores practical implications for English writing pedagogy in Uzbekistan, arguing that NLP-driven AWE — when combined with human oversight — can offer effective, fair, and pedagogically valuable support for large-scale and formative writing assessment.