Language Model for Next Token Prediction (Albanian & English) Description:
This project builds a language model using LSTM to predict the next word based on combined Albanian and English sentences. The model generates coherent text in both languages and evaluates its performance using perplexity. Files:
Albanian.ipynb: Code for training the LSTM model on Albanian and English. model_checkpoint.keras: Best model during training. final_model.keras: Final trained model. loss_history.csv: Training loss over time. Key Outputs:
Generated Text: Generates text in English and French. Perplexity Score: Final perplexity score is 24.58. How to Run:
Clone/download the notebook and files. Install tensorflow, numpy, matplotlib, pandas, keras. Run the notebook to train and generate text.