#ambiguity

Question 1: Single choice

mediumAI-checkedNLP Tasks

In English treebanks only about 14–15% of word types can take more than one part-of-speech tag, yet about 55–67% of the tokens in running text are such ambiguous words. What explains the gap?

The ambiguous words, such as "that", "back", "down" and "set", are very frequent1
Punctuation marks are counted as tokens, and they are usually ambiguous2
Rare words are ambiguous more often, and rare words make up most of the tokens3
The type counts come from a dictionary, while the token counts come from a corpus4

Timed test