Word2Vec and LSTM-Based Deep Learning Model for Lecturer Professional Competence Feedback Classification

Abstract

This study aimed to develop a deep learning model based on Word2Vec and Long Short-Term Memory (LSTM) to classify sentiment in student feedback on lecturers' professional competence. Manual analysis of large volumes of evaluation text data required significant time and resources, thus an automated method was needed to assist the sentiment classification process. Word2Vec was used to represent words as fixed-dimension numerical vectors, which then served as input to the LSTM model. The LSTM model was selected for its ability to process sequential data and retain relevant long-term contextual information in the text. The dataset consisted of 6,124 evaluation texts, divided into 3,800 positive and 2,324 negative samples. The dataset was split into training (70%), validation (10%), and testing (20%) subsets. The model was trained for 50 epochs, achieving a training accuracy of 81.20% and a validation accuracy of 77.10%. Evaluation using a confusion matrix on the testing data showed that the model correctly classified 587 positive and 359 negative samples, while producing 106 false positives and 173 false negatives. These results indicated that the combination of Word2Vec and LSTM was effective in classifying sentiment in lecturer competence evaluation texts, with a testing accuracy of 77.2%.