An Efficient Probabilistic Deep Learning Model for the Oral Proficiency Assessment of Student Speech Recognition and Classification

Abstract

Natural Language Processing is a branch of artificial intelligence (AI) that focuses on the interaction between computers and human language. Speech recognition systems utilize machine learning algorithms and statistical models to analyze acoustic features of speech, such as pitch, duration, and frequency, to convert spoken words into written text. The Student English Oral Proficiency Assessment and Feedback System provides students with a comprehensive evaluation of their spoken English skills and offers tailored feedback to help them improve. It can be used in language learning institutions, universities, or online platforms to support language education and enhance oral communication abilities. In this paper constructed a framework stated as Latent Dirichlet Integrated Deep Learning (LDiDL) for the assessment of student English proficiency assessment. The system begins by collecting a comprehensive dataset of spoken English samples, encompassing various proficiency levels. Relevant features are extracted from the samples, including acoustic characteristics and linguistic attributes. Leveraging Latent Dirichlet Allocation (LDA), the system uncovers latent topics within the data, enabling a deeper understanding of the underlying themes present in the spoken English. To further enhance the analysis, a deep learning model is developed, integrating the LDA topics with the extracted features. This model is trained using appropriate techniques and evaluated using performance metrics. Utilizing the predictions made by the model, the system generates personalized feedback for each student, focusing on areas of improvement such as vocabulary, grammar, fluency, and pronunciation. Simulation mode uses the native English speech audio for the LDiDL training and classification. The experimental analysis stated that the proposed LDiDL model achieves an accuracy of 99% for the assessment of English Proficiency.