Publication

Abstract : This project deals with the testing of different word embedding, deep learning techniques, and XAI tools for text classification tasks. It talks about the effect of Word2Vec embeddings in CNN models on text context and shows their capability of grasping it. Furthermore, this research sees the application of BERT and RoBERTa embedding and shows their influence on CNN and BiLSTM models and their comparable performance. Besides, this project also proves the significance of transformer-based embeddings in the information extraction from texts with various contextual nuances and points to their possible to increase the text classification accuracy on different deep learning architectures. We lastly, look into the function of eXplainable AI tools in the interpretation of model decisions, which ultimately helps to understand and support the model prediction outcomes in the field of text classification.