Frog mblem crash: folia::ValueError: attribute 'class' is required for lemma (empty class passed) about frog HOT 3 CLOSED

Well, this is 'interesting'
In a few special cases the lemmatizer uses tokenizer results to get a lemma.

For instance for the Uppercased word : "In't"
The tokenizer identifies this as a WORD-WITHSUFFIX, and knows how to strip the suffix to get a lemma.
But in this case there are 2 texts associated with the word:

          <w xml:id="vrie047wond01_01.TEI.2.text.body.div.p.note.5800.s.1.w.1" class="WORD-WITHSUFFIX">
            <t>In't</t>
            <t class="contemporary">In</t>
          </w>

AND: we are processing the textclass 'contemporary'
The 'stripping suffix' rule bluntly strips the last 2 characters from "In", leaving an empty word.
So that is wrong and can be improved on.

But there is a fundamental problem: the tokenization was done on the 'current' textclass, but this information is not available. Or we must agree that is ALWAYS on the the 'current' class only. But is NOT what we want.
Should words also carry a textclass? which in this case would mean that we don't have words at al!?
Frog should then extract 'contemporary' sentences from the FoLiA and tokenize those.
A lot of issues arise from here...

Help/thoughts welcome

from frog.

The hack worked fine for my batch, the more fundamental problem is addressed in proycon/folia#32

from frog.

As a note: This is now solved more fundamentally by implementing the textclass attribute for words

from frog.