I AM A STUDENT

-1.5047 I       -1.2566

-2.1363 I AM 
-1.3842 AM A 
-2.8919 A STUDENT
-1.5866 STUDENT IN 

-5.6413 USA     -0.4314
-1.9851 IN      -0.6449



perplexity = exp(-sum_log_prob/(total_words-excluded_ccs-excluded_unks) * 
                log(10.0)),
entropy = (-sum_log_prob/(total_words-excluded_ccs-excluded_unks) * 
             log(10.0) / log(2.0)));


isip18_[1]:../cmulm/evallm -arpa cmu_wb.lm
Reading in language model from file cmu_wb.lm
Reading in a 2-gram language model.
Number of 1-grams = 19126.
Number of 2-grams = 234649.
Reading unigrams...

Reading 2-grams...
...........
Done.
evallm : perplexity -text test_data.text

(1) I 

Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 31.97, Entropy = 5.00 bits
Computation based on 1 words.
Number of 2-grams hit = 0  (0.00%)
Number of 1-grams hit = 1  (100.00%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.


perplexity = exp(1.5047*log10) = 10^1.5047
entropy = 1.5074 *log10 /log 2
log(10.0) / log(2.0) = 2.30259 / 0.69315 = 3.32192


(2) I AM

Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 66.15, Entropy = 6.05 bits
Computation based on 2 words.
Number of 2-grams hit = 1  (50.00%)
Number of 1-grams hit = 1  (50.00%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.

sum_log_prob / 2 = 3.6437 / 2 = 1.82185
perplexity = 10^1.82185 = 66.35



(3) I AM A
evallm : perplexity -text test_data.text
Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 47.32, Entropy = 5.56 bits
Computation based on 3 words.
Number of 2-grams hit = 2  (66.67%)
Number of 1-grams hit = 1  (33.33%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.


sum_log_prob / 3 = 5.0252 / 3 = 1.6751
perplexity = 10^1.67507 = 47.32


(4) I AM A STUDENT


Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 95.34, Entropy = 6.58 bits
Computation based on 4 words.
Number of 2-grams hit = 3  (75.00%)
Number of 1-grams hit = 1  (25.00%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.

sum_log_prob /4 = 7.9171 /4 = 95.34


(6) I AM A STUDENT IN

Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 79.57, Entropy = 6.31 bits
Computation based on 5 words.
Number of 2-grams hit = 4  (80.00%)
Number of 1-grams hit = 1  (20.00%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.

sum_log_prob /5 = 9.5037 /5 = 1.9007
perplexity = 10^1.9007 = 79.568


(7) I AM A STUDENT IN USA


evallm : perplexity -text test_data.text
Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 428.20, Entropy = 8.74 bits
Computation based on 6 words.
Number of 2-grams hit = 4  (66.67%)
Number of 1-grams hit = 2  (33.33%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.


sum_log_prob /6 = 15.7899 / 6 = 2.63165

perplexity = 10^2.63165 = 438.2


(8) I AM A STUDENT IN CHINA


evallm : perplexity -text test_data.text
Computing perplexity of the language model with respect
   to the text test_data.text
Perplexity = 152.12, Entropy = 7.25 bits
Computation based on 6 words.
Number of 2-grams hit = 5  (83.33%)
Number of 1-grams hit = 1  (16.67%)
0 OOVs (0.00%) and 0 context cues were removed from the calculation.


-3.5895 IN CHINA 

sum_log_prob /6 = 2.1822

perplexity = 152.12


(9) IN USA

-1.9851         IN        -0.6449
-5.6413         USA     -0.4314

IN USA  -1.9851*0.4314 0.85637