Your browser doesn't support javascript.
Transfer Learning using BERT & Comparative Analysis of ML Algorithms for Opinion Mining
3rd IEEE Global Conference for Advancement in Technology, GCAT 2022 ; 2022.
Article in English | Scopus | ID: covidwho-2191777
ABSTRACT
Sentiment Analysis is an ongoing field of research in text mining that is concerned with the computational treatment of textual views, sentiments, and subjectivity. It's the task of distinguishing between positive and negative viewpoints, emotions, and assessments. Sentiment analysis has been the topic of intensive research since its inception. During COVID-19, there has been a subtle increase in the usage and the time spent on the social networking sites by people as most of the daily operations have moved online. Moreover, in addition to the illness itself, the pandemic has led to dread, anxiety, stress, concern, repugnance, and poignancy in individuals all around the world. Considering these indicators, experts are paying close attention to Twitter data analysis during this pandemic. BERT is used as a transfer learning model and this work analyses the efficacy of fine-tuning it for the task of opinion mining by comparing it to a baseline model that includes a TF-IDF vectorizer and a Naïve Bayes classifier. Its performance is also compared to that of Naïve Bayes, Logistic Regression, K Nearest Neighbor, Decision Tree and XGBoost classifier. To determine the most effective settings for the BERT model, hyper-parameter tweaking is used. After two epochs of training at a learning rate of le-5 and batch size of 16, the maximum accuracy of 87.6% is attained. These results outperform all of the machine learning models examined in this study. This work tackles a comprehensive overview of the last update in this field. It can be beneficial to scholars in this domain because it encapsulates the most well-known Sentiment analysis methodologies and their comparison in single research work. © 2022 IEEE.
Keywords

Full text: Available Collection: Databases of international organizations Database: Scopus Language: English Journal: 3rd IEEE Global Conference for Advancement in Technology, GCAT 2022 Year: 2022 Document Type: Article

Similar

MEDLINE

...
LILACS

LIS


Full text: Available Collection: Databases of international organizations Database: Scopus Language: English Journal: 3rd IEEE Global Conference for Advancement in Technology, GCAT 2022 Year: 2022 Document Type: Article