| I developed a novel automatic hip-hop lyrics generation model by integrating image recognition techniques with linguistic generation models. The model, based on VisionEncoderDecoderModel and GPT-2, converts image content into hip-hop lyrics in the style of the featured artist. Through pre-training the image recognition model and fine-tuning the language model with artist-specific lyrics datasets, the model learns language patterns and rhyme structures. Optimization techniques such as specific processing functions and an improved loss function ensure rhyme density, coherence, and stylistic consistency in the generated lyrics. Objective and subjective evaluations validate the model’s performance, advancing automated hip-hop lyrics generation and exploring innovative intersections of music and visual arts. |
