Comparing the performance of various Encoder Models and Vectorization Techniques on Text Classification

This paper presents a comprehensive comparison of the performance of various encoder models and vectorization techniques for text classification tasks. The study includes an evaluation of various pre-trained Autoencoding Transformers such as BERT, DistilBERT and Roberta, as well as traditional machine learning models used along with vectorization techniques such as BOW and TF-IDF. The authors conducted experiments on a diverse set of benchmark datasets from the Kaggle problem statement to classify disaster tweets as catastrophic and non-catastrophic and compared the accuracies of different models, computational efficiency, and scalability. The results of this study provide a valuable reference for researchers and practitioners looking to choose the best encoder models and vectorization techniques for text classification tasks.