Tools
Tools
Sharma, Daksh (2023) Opinion Mining using Twitter data for Ukraine-Russia War. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (530kB) | Preview |
Abstract
The conflict between Ukraine and Russia (RUW) escalated in 2022 but has been a topic of interest since 2014. Social media platforms like Twitter contain raw data that can be utilised to gain knowledge about public opinion and attitude towards a particular topic. In this study, tweets relating to the Ukraine-Russia conflict were obtained and used to provide insights about public opinion in the year 2023. The results were compared to the results obtained in previous studies until 2022 to dig deeper into the public opinion and figure out whether the sentiments of the people have shifted or remained constant. The main objective of this study is to effectively utilize BERT model to perform opinion mining on tweets relating to RUW. Our approach involves using VADER and RoBERTa to label the tweets and then use the labelled data to fine-tune a BERT model. In the second approach, the results from both VADER and RoBERTa were merged to create a dataset with true sentiment labels, which was used to fine-tune another BERT model. RoBERTa based BERT was found to be performing better than VADER based BERT with an average accuracy difference of 33.44%. This gives us the idea that transformers-based models are more effective in performing sentiment analysis than rule-based approaches.
Actions (login required)
 |
View Item |