Performance Optimization for Transformer Models on Text Classification Tasks

The emergence of social media has led to a new set of celebrities, known as influencers. Influencers are individuals who have built a following on social media platforms such as Instagram, YouTube, and Twitter. These individuals have the power to influence the opinions and purchasing decisions of their followers. As a result, brands have sought out partnerships with influencers to promote their products or services. However, with the rise of sponsored content, it has become increasingly difficult to discern whether a post is genuinely endorsed by an influencer or is paid for by the brand. This paper explores how machine learning models can be used to classify influencer tweets as sponsored or not sponsored. Through the use of transformer language models such as BERT, GPT 2 and GPT Neo, we can identify patterns and relations between an influencer and a brand. By analyzing the content and context of influencer tweets, machine learning models can help identify whether a tweet is a sponsored post or not. This paper also takes into consideration the performance metrics such as training time and draws comparison between the models. The use of machine learning models to classify influencer tweets as sponsored or not sponsored will aid in the development of new metrics that accurately measure the effectiveness of influencer marketing. This will, in turn, lead to a more transparent and mutually beneficial influencer marketing industry.