take the block on the green circle			P = 0.004500
put the block on the circle on the red circle		P = 0.002699
put the green cone on the square			P = 0.003000
take the red cone					P = 0.020000
put the green block on the square			P = 0.006000
take the blue cone on the red circle			P = 0.000375
take the cube on the red circle				P = 0.004500
put the blue cone on the red circle on the green circle	P = 0.000042
put the block on the green circle			P = 0.004500
put the blue block on the circle			P = 0.006000

Probability of corpus: 8.3671 E-27

Entropy (base 2) (per sentence): 8.6627
Entropy (base 2) (per word): 1.1398

Perplexity (per sentence): 405.2681
Perplexity (per word): 2.2035

NB: The test corpus was generated with the same probabilistic grammar
    as the training corpus. The sentence probabilities are derived 
    from the grammar and sum to 1 over the set of all possible strings 
    given the 12 word vocabulary.

    The probability of the corpus was computed simply by multiplying
    the sentence probabilities, which amounts to assuming that sentences
    are probabilistically independent of each other.

    The entropy (per sentence and per word) was estimated simply by
    taking the logarithm (base 2) of the corpus probability and
    dividing by the number of tokens (sentences and words, respectively).
    This empirical estimate of the entropy is sometimes simply
    called the logprob (LP).

    The perplexity is 2 raised to the power of the entropy (i.e., the
    entropy is the logarithm of the perplexity).