Natural Language Processing for Sentiment Analysis in ESL Writing: Understanding Emotional Tone in Learner Essays

In recent years, the significance of sentiment analysis in educational contexts has gained considerable attention, particularly in understanding the emotional expressions of English as a Second Language (ESL) learners. Accurately capturing these sentiments is essential for providing targeted support and feedback to enhance student writing skills. This study presents an advanced natural language processing (NLP) framework for sentiment analysis in ESL writing, specifically focusing on understanding emotional tones in learner essays. Utilizing a dataset sourced from Kaggle, which categorizes tweets based on emoticons, the methodology incorporates rigorous data preprocessing techniques, including text cleaning, tokenization, and error correction to enhance data quality. A comprehensive feature extraction process captures linguistic, syntactic, and sentiment-related features using dependency parsing and word embeddings, such as Word2Vec and BERT. The proposed model, implemented in Python, demonstrates a remarkable accuracy of 98.5%, significantly surpassing existing methods like Recurrent Neural Networks (RNN) and Convolutional Neural Networks (CNN), which achieved accuracies of 95.12% and 96.97%, respectively. This enhanced performance highlights the model's ability to effectively capture nuanced emotional expressions in ESL writing. The validation of the model's predictions against human evaluations ensures the reliability and accuracy of the findings. This research not only contributes to the field of sentiment analysis but also informs educational practices, offering targeted interventions to support ESL learners in improving their emotional expression and writing skills. Ultimately, the proposed NLP framework serves as a robust tool for enhancing understanding and assessment of emotional tone in educational contexts.