Needs Clarification:

Split it in two steps. The second step should only be shown when we can not guess the information in the second step.

I roughly remember this was about the language selection. The first step was selecting the Language of the spoken word, the second was mainly needed if there are two different ways to spell an otherwise equivalent word. And that second step should only be avalible if needed, since otherwise a) the user remains confused b) the user may select wrong things while selecting nothing at all is fine most cases and most languages.

Here a mock of the possible lexeme creation UI

There is an additional optional field, The bottom field, "Language of Lemma", which would be optional and only shown, if the language of the lexeme is one that has several written forms. If you want.

The current UI on which it was based:

Differences:

• Lemma on top (since this is likely to be the most important thing for the user, also we may be able to suggest a language based on it’s input)#
• Language of lemma optional (since it is only needed if there are e.g. different spellings for the same concept)
• Help icons. The wording is very technical/sciency. The least we can do is to provide help.
• Examples for the kind of input in the secondary text in the box instead of the generic and providing-a-technical-access "Enter ID"

Change 341577 had a related patch set uploaded (by jakob):
[mediawiki/extensions/WikibaseLexeme] [WIP] Add item selector to lexeme language textfield.

https://gerrit.wikimedia.org/r/341577

Change 342243 had a related patch set uploaded (by Jakob):
[mediawiki/extensions/WikibaseLexeme] [WIP] Infer lemma language from lexeme language for new lexemes.

https://gerrit.wikimedia.org/r/342243