C1_W1: Erro in frequency count part 2

Course Q&A Natural Language Processing NLP with Classification and Vector Spaces
1

Continuing the discussion from Error in frequency count:

Issues happend in the frequency table generated by build_freqs, when compared my results to another results have the same issues at Apr.11,

Continuing the discussion from Error in frequency count:

i see this had changed. I try to test this function many time, do the same as hints but still didn’t get the right answer. I have not edited the utils.py file and everything up to this point works well, but it does cause me issues later on.

hime
hime770×560 47 KB

mine
mine1014×435 30.7 KB

2

Send me in private the code for extract_features, the error should be there, i will have a look on it, because i have no idea where the error might be otherwise.

3

It is because you have not followed the instructions properly:

Implement the extract_features function.

  • This function takes in a single tweet.
  • Process the tweet using the imported process_tweet function and save the list of tweet words.
  • Loop through each word in the list of processed words
    • For each word, check the ‘freqs’ dictionary for the count when that word has a positive ‘1’ label. (Check for the key (word, 1.0)
    • Do the same for the count for when the word is associated with the negative label ‘0’. (Check for the key (word, 0.0).)

Note: In the implementation instructions provided above, the prediction of being positive or negative depends on feature vector which counts-in duplicate words - this is different from what you have seen in the lecture videos

First the x is not 2 dimensional but 1 dimensional of shape 1x3.

Second, you use word, 1.0- floating number

Third, the batch dimension is added at the end (already done for you) but you removed it.

1 Like
4

thank you so much, i had passed it.
But i code the same images that i send you and no errors appears again, i dont know why :joy:

5

Hi. I’m having similar problem with this part. The testing tweet word list after processing seems to be [‘followfriday’, ‘top’, ‘engag’, ‘member’, ‘commun’, ‘week’, ‘:)’] which has no negative word according to freqs dict. So I got [[1.000e+00 3.133e+03 0.000e+00]]

image
image1669×334 28.3 KB

after [23] test which is different with the expect output [[1.000e+00 3.133e+03 6.100e+01]]
.

I was using tuple[str, float] key to select freqs dict but i think it has nothing to do with the results since there is no negative words in the tweet list, if i didnt make other mistakes.

Thanks for help.

6

I printed out all word records for debug:

image
image1375×586 43.3 KB