Lemma column needed in conllu input?

[jowagner]

The example in the readme shows conllu code with ID, form and lemma column populated. However, all read_conll_*corpus functions in __main__.py do not read the lemma column. Do you read the lemma column anywhere else or do you plan to use it in the near future? Do I need a lemmatiser?

Of course, udpipe users can get the lemmatisation from udpipe and one will want to use udpipe with the offered models as udpipe's tokenisation differs from other popular tokenisers. The lemma column would, however, be a difficulty if I use my own tokeniser and train my own models (as described in the readme).