chunked/nltk_chunking/_perplexity/chunk_13.txt

When evaluating the model's perplexity of a
sequence, a tempting but suboptimal approach is to break the sequence into disjoint chunks and add up the decomposed
log-likelihoods of each segment independently.