Lexical Bias In Essay Level Prediction

TL;DR

This paper introduces 'balikasg', a system that achieved state-of-the-art results in predicting non-native English essay levels, emphasizing feature engineering and model choices.

Contribution

It presents a novel system with detailed feature extraction and engineering strategies that outperform previous approaches in essay level prediction.

Findings

Achieved state-of-the-art performance in CAp 2018 challenge

Analyzed impact of feature engineering on accuracy

Provided insights for future improvements

Abstract

Automatically predicting the level of non-native English speakers given their written essays is an interesting machine learning problem. In this work I present the system "balikasg" that achieved the state-of-the-art performance in the CAp 2018 data science challenge among 14 systems. I detail the feature extraction, feature engineering and model selection steps and I evaluate how these decisions impact the system's performance. The paper concludes with remarks for future work.