Our project is focused on generating lyrics inspired Ed Sheeran. We identify recurring themes, patterns, and structures in his songwriting through multiple algorithms.
Our primary objective is to develop text generation models using Language Model such as N gram, LSTM and Transformer. Additionally, we leverage these models to craft new lyrics in the same style as Ed Sheeran. Model Evaluation is done carefully in order to examine and contrast these lyric-generating language models and determine the most effective approach.
The Ed Sheeran dataset comprises 179 rows and 3 columns, labeled "Song name," "Lyrics Link," and "Lyrics Content." It includes 111 unique songs, with a vocabulary size of over 4000 unique words across all the lyrics. The dataset contains a total of 62,000 plus tokens, representing the total number of words used in the lyrics.
Ed Sheeran Lyrics generator:
Steps to run the program:
- install all the requirements in the requirements.txt - pip install -r requirements.txt
- Run the cells in each file
Drive Link: https://drive.google.com/drive/u/2/folders/1yUIx4jULrkecdizwBv9vmTETPxeXY415
Recording Link: https://docs.google.com/document/d/1HSOoK4W86cpVhLt2JA5oJhIDytICkd18cgOhNxNJpBY/edit?usp=drive_link
Report Link: https://drive.google.com/file/d/18YAg6l7EzHRYhTdHq7hYq8XGK9p_z6Re/view?usp=drive_link
Proposal Link: https://drive.google.com/file/d/1x_hcdLqvp-I11Ec3aEbY7ZDfXSVdhwZl/view?usp=drive_link
GIT Repository Link: https://github.com/sriksven/NLP_Project_LyricLab
Link to Dataset: https://www.kaggle.com/datasets/kedar123/ed-sheeran