UCLA

This research paper explores the application of Transformer models in the field of Natural Language Processing (NLP) to detect sponsored content in Spotify audio data. The paper discusses the evolution of Transformers in NLP, highlighting their efficiency and accessibility in creating meaningful narratives from large datasets. The study focuses on a dataset of 100,000 Spotify Podcasts and their descriptions, aiming to achieve three objectives: Classification, Named Entity Recognition, and Topic Modeling. The research utilizes Transformer models, specifically through the Hugging Face library, to fine-tune and implement state-of-the-art models for efficient analysis. The application of Transformer models, including BERT, promises to save time and resources compared to traditional methodologies.