can Text only appear in tags named TextLine?

[gundramleifert]

so far I assume to have text only in these structures. Is that right?
<TextLine> <TextEquiv> <Unicode> TEXT </Unicode> </TextEquiv> </TextLine>

[kahlep]

Complicated topic ;-)

• If PAGE XML has been imported into Transkribus there may be also text in */TextRegion/TextEquiv.
  However, this is currently not visible/ editable in Transkribus GUI. So this can be savely ignored for now. But that mode might be needed for text2image matching!?

• There is a mode for word-based edition (with polygonal word regions). E.g. OCR output is transformed to a PAGE XML using all levels of fulltext, including this one. Then a user might correct only on word level, and optionally sync the corrected text to line level, which should be done for correctness but is up to the user.

We do not store the edit-mode a user has been using on a page. We discussed that a lot but did not come to a conclusion yet.

So my opinion on this is:
an optional parameter (that defaults to line-level) for specifying where the text has to be extracted from might come in handy later. As error rate is mostly computed on HTR output and raw OCR output, I think we can stick to line-level for now.