word segmentation in kant_aufklaerung_1784 GT PageXML about assets

[wrznr]

@bertsky Yes, such a strict rule would be better. Since we do not have a designated space element as in ALTO and no wordStart attribute as in Abbyy XML, I would even say that this is the only possible way to unambiguously represent whitespace.

[wrznr]

@tboenig assets and our own GT has to be thoroughly checked wrt this issue.

[bertsky]

Excellent! Maybe having an automatic validation step (see #16) for consistency between PAGE element levels would quickly reveal such non-abiding annotations. (Or even allowed exceptions to it in corner cases like extra whitespace at paragraph column crossings, drop-capital regions, etc. So one might need a semi-automatic procedure to get there.)

[wrznr]

@tboenig Push.

[wrznr]

@tboenig Push.

[bertsky]

@tboenig Please reopen. I understand that this is allowed now by the spec under a "lax" consistency rule. In fact, all variants kant_aufklaerung_1784{,-jp2,-binarized,-page-block-line-word_glyph} fall under this case.

But can we have at least one representation here in the GT assets repo that does meet the strict rules for tokenisation? (We need this to be able to write automated module tests without having to download GT files each time ...)

[tboenig]

We made a new example this month.

[bertsky]

In the assets repo? Please publish!

[kba]

I think @tboenig meant we WILL provide a new example in the coming weeks.