Data Science and Intelligent Computing Techniques
An Opinion Mining-Based Pretrained Model Analysis Depending on Multiple Thinking Patterns
Authors: Joydhriti Choudhury, Abdullah Al Farabe, Adria Binte Habib and Muhammad Iqbal Hossain
Publishing Date: 05-10-2023
ISBN: 978-81-955020-2-8
Abstract
People love to express their opinions online, and the use of social media for this purpose has skyrocketed in the Internet age. Customers also frequently share their first-hand knowledge of the goods or services they have used. Reviews can be either favorable or bad, and both have an impact on people's decisions and businesses. Therefore, it is essential to predict people's opinions in order to preserve and understand the reliability of online review systems. Our study concentrates on a certain kind of model, namely an opinion mining-based pre-trained model, and its analysis takes different types of thought patterns into account. Pretrained models based on transformers, such as BERT, RoBERTa, and DistilBERT may be effective at classifying people's opinions. Other classifiers, such as DISTILBERT and RoBERTa, offer more accuracy compared to the rest of the classifiers employed by researchers but in our model, RoBERTa offers the best accuracy at 90%.
Keywords
EDA, tokenization, BERT, RoBERTa, DistilBERT
Cite as
Joydhriti Choudhury, Abdullah Al Farabe, Adria Binte Habib and Muhammad Iqbal Hossain, "An Opinion Mining-Based Pretrained Model Analysis Depending on Multiple Thinking Patterns", In: Satyasai Jagannath Nanda and Rajendra Prasad Yadav (eds), Data Science and Intelligent Computing Techniques, SCRS, India, 2023, pp. 317-327. https://doi.org/10.56155/978-81-955020-2-8-28