UNQ_C8 incorrect, any change in the corpus?

Course Q&A Natural Language Processing NLP with Probabilistic Models
1

I don’t think my work has problem. Why it’s incorrect? I guess if the corpus is changed, the expected output should be adjusted as well.

# UNQ_C8 (UNIQUE CELL IDENTIFIER, DO NOT EDIT)
# GRADED FUNCTION: compute_accuracy
def compute_accuracy(pred, y):
    '''
    Input: 
        pred: a list of the predicted parts-of-speech 
        y: a list of lines where each word is separated by a '\t' (i.e. word \t tag)
    Output: 
        
    '''
    num_correct = 0
    total = 0
    
    # Zip together the prediction and the labels
    for prediction, y in zip(pred, y):
        ### START CODE HERE (Replace instances of 'None' with your code) ###
     # code removed - not allowed by the code of conduct
        ### END CODE HERE ###
    return num_correct/total

Capture22_3
Capture22_3920×416 43.4 KB

2

If you print out word_tag_tuple, you can see it contains noises. I used strip to remove unwanted characters before the split, and split is based on ‘\t’

4

I think it’s because you use tuple() before checking if len(word_tag_tuple)!=2. You manually added a lot of cases not equal to the length of 2.

5

Thanks for your reply.
However, I don’t think tuple() function will change original length. Actually I followed youe advice and change the line to
word_tag_tuple = y.strip().split(’\t’). I got the same result as before. Can I ask in your work how many num_correct did you get in your work?

1 Like
6

Well, sorry, in that case, I don’t see this part has problem…
my output is:
num_correct: 31311
total: 32853
Accuracy of the Viterbi algorithm is 0.9531

7

I’m having a similar issue, and stripping out whitespace or where I cast to a tuple (via tuple()) doesn’t change the result.

I get the expected accuracy (0.9531%), but when I submit the assignment, I get 0/10 for the Compute Accuracy step, and the grader output just says:

There was a problem grading your submission. Check stderr for more details.

I don’t think there’s actually a way to inspect the grader’s traceback or to check the stderr. If there is a way, I’d love to hear about it!

8

I am facing the same issue. i got .9528 accuracy.
how did you get .9531 accuracy?

9

I am getting an accuracy of 0.9168.
I see that another person on this list also got this. Was this issue ever resolved? i.e. how to get 0.9528?

My “viterbi_backward” → method has some issue. Is the above problem of 0.9168 due that? I.e. unless you get “Viterbi_backward” correct, you can’t get the method “compute_accuracy” correct?

10

The problem is the ‘pred’ parameter in the ‘compute_accuracy’. You should check UNQ_Q7 again despite “All tests passed” appeared. Hope it help.

2 Likes
11

Hi, I am getting the stated 0.9531 accuracy instead of 0.9528 if I give correctly the pos_tag_for_word_i variable in viterbi-backward function as per instruction