I'm working on human evaluation/annotation of text generation models and am wondering if there's a way to use a
text_input field that is pre-filled with the text to be annotated.
The idea is similar to the image captioning demo example, but instead of asking the annotator to type the caption from scratch, we want to provide an editable text. This should allow the annotator to quickly edit words in place with the cursor rather than having to re-type or manually copy/paste the parts of the text that are ok.
The gaol here would be to capture edit operations between a model's output and a human-edited version.
Thanks in advance!