Anaphraseus eats footnotes
The easiest solution would be to support SRX format for sentence segmentation: https://en.wikipedia.org/wiki/Segmentation_Rules_eXchange...
Give more detailed error message when loading failed
More data about corpora (version number and build date)
stop the current search
search on several corpora and save the selection
Isnt, Doesnt, Hasnt... Etc