Comparison of Word Embeddings for Sentiment Classification with Preconceived Subjectivity
Xi Jie Lee, Timothy Tzen Vun Yap, Hu Ng, Vik Tor Goh · 2022
This research looks into objectivity and subjectivity's effects on sentiment analysis through word embeddings, namely Word2Vec, Term Frequency-Inverse Document Frequency (TF-IDF), and Bidirectional Encoder Representations from Transformers (BERT).Objectivity corpora are defined as data with a neutral point of view and no biases.In contrast, subjectivity corpora are defined as data from a non-neutral point of view and may contain biases.The goals are to compare the efficacy of numerous embedding methods on sentiment analysis classification after subjectivity analysis.In terms of embedding methods, results from our work show that BERT embedding gives the best outcome for subjectivity classification with an accuracy score of 99.77%.For sentiment classification, TF-IDF provides the highest accuracy with 91.29%.