#machine-translation

Question 1: Single choice

A translation dataset builds its source and target vocabularies from the first max_examples sentence pairs, keeping only the words that occur at least twice. What happens to the two vocabulary sizes as you raise max_examples?

Both grow or stay the same: new words appear and more rare words reach a count of 21
Both shrink, because the counts are spread over more and more sentence pairs2
Only the source side grows; the target side is fixed by the size of its alphabet3
Neither changes, because the threshold of 2 already fixes the vocabulary size4

#machine-translation

Timed test