This project presents an end-to-end implementation of a Part-of-Speech (POS) tagging model using a Long Short-Term Memory (LSTM) neural network.

POS tagging involves assigning grammatical categories (such as noun and verb) to words in a given text. The model is developed using PyTorch and is trained on a dataset derived from the Penn Treebank.

The primary data source was the Billboard Year-End Hot 100 Singles USA, an annual ranking of the top-performing songs in the United States. It encompasses 5402 lyrics out of the 6889 songs from a diverse array of artists and genres.

The dataset spans from 1946 to 2022, capturing the evolution of popular music over several decades. Notably, the first years of the ranking included less than 100 songs.

To ensure a representative dataset, at least the top 20 songs from each year were included. This criterion was applied to account for variations in the number of songs listed each year.