[nescioquid]

One suggestion for reducing the burden on your users would be to start making predictions about what vocabulary your user already knows, just based on what's known about word frequency and a short quiz pulled from the text.

I copied in a long poem and it looked like I was going to be prompted about whether I knew 1500+ "words" (are you lemmatizing the input at all, BTW?). If your user knows the most common verbs, they probably already know the prepositions, pronouns, and other closed lexical classes of words (and vice versa). If your user is familiar with less common vocabulary (e.g. something at C1), raise the word frequency threshold for checking if the user is familiar with the word. If your user is less familiar with basic vocabulary, don't overwhelm them with moderate and advanced vocabulary.

That would make the prompting portion more interesting -- you select the most discriminating words to zero in on estimating the user's ability (this is really how adaptive testing works). You could gamify this too, by essentially establishing the user's "vocab ELO" rating based on word frequency.

Admittedly, maybe my suggestion misses the point of your app in case the objective was to be sure that you don't miss any new-to-you vocab in a text. On the other hand, if you could do something along the lines of my suggesting, you won't overwhelm beginners and you won't exasperate more advanced learners.