Sequential Models for Sentiment Analysis: A Comparative Study

• Published in: Advances in Computational Intelligence Vol. 13613; pp. 227 - 235
• Main Authors: Adebanji, Olaronke Oluwayemisi, Gelbukh, Irina, Calvo, Hiram, Ojo, Olumide Ebenezer
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer 2022; Springer Nature Switzerland
• Series: Lecture Notes in Computer Science
• Subjects: Deep learning algorithm; Machine learning algorithm; Sentiment analysis; Sequence modeling; Word embedding
• ISBN: 9783031194955; 3031194950
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-031-19496-2_17

Sentiment analysis has been a focus of study in Natural Language Processing (NLP) tasks in recent years. In this paper, we propose the task of analysing sentiments using five sequential models and we compare their performance on a Twitter dataset. We used the bag of words, as well as the tf-idf, and the Word2Vec embeddings, as input features to the models. The precision, recall, f1 and accuracy scores of the proposed models were used to evaluate the models’ performance. The Bi-LSTM model with Word2Vec embedding performs the best against the dataset, with an accuracy of 84%.