chunked/nltk_chunking/_perplexity/chunk_4.txt

Intuitively, it can be thought of as an evaluation of the model's ability to predict uniformly among the set of specified tokens in a corpus.