Development of a Nigerian English classification model: for AI-Driven Grading Systems

Abstract

The current Automated Essay Assessment Systems (AEAS) are predominantly trained on native Standard English, thereby introducing bias when grading essays written in other variations of English- Nigerian English. This bias leads to unfair grading, misclassification of valid linguistic features, and an increased failure rate among students. Nigerian English, the official language of Nigeria, incorporates linguistic features that differ from native English expressions. This study aims to enhance grading fairness by developing a Nigerian English classification model using K-Nearest Neighbors (KNN) and Term Frequency-Inverse Document Frequency (TF IDF). The model successfully identifies and classifies Nigerian lexical features, by incorporating Nigerian English dictionaries and crowdsourced speech resources, aiding in unbiased assessments. Results suggest that this approach significantly improves recognition of Nigerian English expressions, contributing to fairer academic evaluations.