Harnessing AI and Machine Learning to Transform English Language Learning Experiences

This study presents a methodology for evaluating argumentative essays from English language learners (ELLs) in grades 8 through 12, utilizing the ELLIPSE corpus. The essays are assessed across six critical metrics: coherence, syntax, vocabulary, phraseology, grammar, and conventions, with ratings from 1.0 to 5.0. The process begins with data collection, followed by preprocessing steps to prepare the text for machine learning. This includes removing punctuation, stopwords, stemming, lemmatization, and expanding contractions to enhance text quality for analysis. Advanced feature extraction techniques are then applied, utilizing word embeddings to convert textual data into dense vector representations. These embeddings capture the syntactic and semantic nuances of the text, making it suitable for machine learning models. The vectorized data is input into Long Short-Term Memory (LSTM) networks, which are adept at handling sequential data. LSTMs manage information flow through cell states and gates, allowing them to capture long-term dependencies in the text effectively. The LSTM network processes the text sequentially, leveraging previous words to make informed classification decisions. The resulting model provides high-accuracy classifications of essays based on the six metrics, offering insights into learners' proficiency levels. This methodology demonstrates a robust application of AI and machine learning in enhancing English language learning, with potential uses in personalized education and automated assessment systems. The study highlights the effectiveness of integrating advanced feature extraction and sequential modeling techniques to assess and improve language proficiency among students. The proposed Word Embedding-based LSTM Model, however, gave out a great result where it produced an accuracy of 98% and it is implemented in python.