Μοντέλα transformer για την ταξινόμηση στιχουργών
Transformer models for song lyrics writers classification
View/
Keywords
Artificial intelligence ; AI ; NLP ; Classification problemAbstract
This thesis focuses on the architecture of deep learning models (transformers). It analyzes the architecture and presents the implementation of this specific model on song lyrics to classify each line with its corresponding artist who created it.
Initially, related approaches to text data problems that utilize the transformer architecture as a model are examined. Furthermore, the necessary data preprocessing steps before model training are described, along with an analysis of the data to ensure proper training process.
Next, the transformer architecture is elaborated in detail to provide an understanding of how these models can achieve results similar to human-like performance. Visual illustrations are used at several points to facilitate comprehension of the architecture.
Then, a detailed description of the training data used for the model is provided, along with the preprocessing steps required prior to training.
Finally, the thesis concludes with the final findings and areas for further research.