Sentimental Analysis of COVID-19 Vaccine Tweets Using BERT+NBSVM

Improving Healthcare Communication: AI-Driven Emotion Classification in Imbalanced Patient Text Data with Explainable Models

“…(2) Furthermore, compared to works based on artificial learning [8][9][10][11][12][13], this study not only improves the transparency of results by integrating explainability into analysis but also contributes to the growing field of explainable AI in sentiment analysis. This approach not only advances this understanding of sentiment classification in a critical area but also sets a precedent for future research in applying explainability to deeply understand model decisions.…”

“…Das et al [12] compared different deep learning and hybrid models for the sentiment analysis of comments on an e-commerce site, for the classification of texts in English and Bengal, and demonstrated that the model support vector machine (SVM) outperformed other models, achieving an accuracy of 82.56% for sentiment analysis of English texts and 86.43% for sentiment analysis of Bengali texts. Umair et al [13] proposed an approach to analyze tweets related to COVID-19 vaccines and combined the BERT + NBSVM model to classify people's sentiments towards vaccines. This choice is motivated by taking advantage of both bidirectional BERT and NBSVM functionalities from transformers and circumventing the limitations of BERT-based approaches, which only leverage encoder layers, resulting in lower performance on short texts.…”

“…Given the impressive results of deep learning algorithms, particularly in the analysis of textual data [11][12][13], an exploration of their scope on the EmoHD dataset, so two types of recurrent neural networks (RNN) are implemented: LSTM and the BERT, and compared them to the learning algorithms implemented previously based on their classification performance.…”

“…129-142 Based on these expertise, this study focuses on the EmoHD dataset to demonstrate the practical application of sentiment analysis techniques in a healthcare context. This study framework therefore differs from the works cited [8][9][10][11][12][13], as its focus is on applying various sentiment analysis methodologies to the EmoHD dataset, which contains a unique compilation of textual data. derived from patient feedback on serious illnesses.…”

“…By leveraging insights from the EmoHD dataset, this study (in comparison with [4,14]) aims to contribute further by evaluating the performance of different learning algorithms such as logistic regression, Bayes naive, and LightGBM. integrating a variety of metrics (CBOW, TF-IDF, and Word2vec) and comparing them to recurrent deep learning networks, including LSTM and BERT models which have shown good results in sentiment classification [10][11]13].…”