Future Work

📘 Future Work

This project lays the foundation for BERT+GPT-2 based text simplification. Future improvements could include:

🔹 Model Enhancements

Explore other encoder-decoder models (BART, T5, LLaMA-based).

Use reinforcement learning to directly optimize for SARI instead of cross-entropy.

Integrate control tokens to adjust simplification level (mild, moderate, strong).

🔹 Data Improvements

Expand training data using synthetic simplifications (back-translation, paraphrasing).

Domain-specific simplification datasets (medical, legal, education).

🔹 Applications

Build a web interface or API to test simplification interactively.

Integrate into educational tools for children or non-native speakers.

Use in assistive technologies (reading aids, accessibility software).