Multi-language form support

[dogancanbakir]

The current model is trained primarily on English-language forms. Forms in other languages (labels, placeholders, button text) may not classify as accurately.

Explore training on multilingual datasets to improve coverage for non-English forms.