There is a lot of room to improve this project, so this folder contains some initial progress we made in extending our work. We've also compiled other ideas for how to contribute:
- characterize the edge cases missed by the constraint satisfaction problem used to pair acronyms with their ground truth definitions
- retrain model with more acronyms
- optimize GloVe feature vector accuracies by increasing the size of vectors or selected context words.
- implement calls to existing acronym database (via Stands4 API) for classifying unseen acronyms
- potentially using the category of the surrounding text to better hone in on the correct definition
- host Chrome extension
- update Chrome extension to accept user feedback about the validity of the acronym suggestion