Answer span extraction in NLP - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to extract the start index of the answer span from model outputs.

NLP

start_index = outputs.start_logits.[1]()

Drag options to blanks, or click blank then click option'

Amax

Bargmax

Csum

Dmean

Attempts:

3 left

💡 Hint

Common Mistakes

Using max() returns the value, not the index.

Using sum() or mean() does not give the position.

The argmax function returns the index of the highest value, which corresponds to the predicted start position of the answer span.

2fill in blank

medium

Complete the code to extract the answer text from the context using predicted start and end indices.

NLP

answer = context[[1]:end_index + 1]

Drag options to blanks, or click blank then click option'

Astart_index

Boutputs.start_logits

Cend_index

Doutputs.end_logits

Attempts:

3 left

💡 Hint

Common Mistakes

Using logits instead of indices.

Using end_index as the start of the slice.

The answer span is extracted from the context starting at start_index up to end_index inclusive.

3fill in blank

hard

Fix the error in the code to correctly compute the end index of the answer span.

NLP

end_index = outputs.end_logits.[1]()

Drag options to blanks, or click blank then click option'

Aargmax

Bmax

Cmin

Dsum

Attempts:

3 left

💡 Hint

Common Mistakes

Using max() returns the value, not the index.

Using min() or sum() is incorrect for this purpose.

Using argmax returns the index of the highest end logit, which is the predicted end position of the answer span.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps tokens to their indices for tokens longer than 3 characters.

NLP

{ [2]: [1] for [1], [2] in enumerate(tokens) if len([2]) > 3 }

Drag options to blanks, or click blank then click option'

Aindex

Btoken

Ci

Dword

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'i' instead of 'index' for the value.

Using 'word' instead of 'token' in the loop.

The dictionary comprehension maps each token to its index. The loop variable is token from enumerate(tokens).

5fill in blank

hard

Fill all three blanks to create a dictionary of token lengths for tokens starting with a vowel.

NLP

{ [1]: len([2]) for [3] in tokens if [2][0].lower() in 'aeiou' }

Drag options to blanks, or click blank then click option'

Atoken

Bword

Cw

Attempts:

3 left

💡 Hint

Common Mistakes

Mixing variable names in keys and values.

Using different variable names inconsistently.

The dictionary comprehension uses token as the key and calculates the length of token. The loop variable is also token.