A MULTIFACETED SENTIMENT ANALYSIS APPROACH TO THE ESTIMATION OF THE STRENGTH OF ONLINE SUPPORT FOR POLITICAL CANDIDATES IN NIGERIA'S ELECTIONS

The strength of online support for political candidates in an election is crucial to their victory at the polls, particularly in countries with advanced digital infrastructure and culture. In modern times, social media is one free space where residents express, and are persuaded to, support or show disdain for political candidates prior to an election. This has resulted in the opinion mining of political tweets to predict electoral victories at the polls. However, this is usually done by adopting a single sentiment analysis model and scraping tool. Ordinarily, no sentiment analysis model or scraping tool is a silver bullet – each has strengths and weaknesses. Thus, this study employed two contemporary scraping tools and adopted three contemporary sentiment analysis models. The models were then exposed to the scrapped political tweets of the top contestants for the Nigeria 2023 presidential election, validated with another set of political tweets of the top contestants for the 2024 Edo State governorship election, and, after that, used to predict the online support strength of the top contestants for the 2024 Ondo State governorship election. Only tweets from within the geopolitical space of elections were scrapped. A notable finding of this study is that no two sentiment analysis models estimate the same online support strength for selected candidates, even with the same set of tweets. Overall, the study holds that online support strength is necessary but insufficient to guarantee victory at the polls in Nigeria's elections.