Text chunking strategies in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why use text chunking in language models?

Which of the following best explains why text chunking is important when processing long documents with language models?

A It helps break down long texts into smaller parts so the model can process them without losing context due to input length limits.

B It increases the total number of words in the text to improve model accuracy.

C It removes all punctuation to simplify the text for the model.

D It translates the text into multiple languages before processing.

💡 Hint

Think about model input size limits and how chunking helps manage them.

Language models have limits on how many words or tokens they can process at once. Chunking splits long texts into smaller pieces so the model can handle them effectively without losing important information.

❓ Predict Output

intermediate

Output of text chunking code

What is the output of this Python code that chunks text into parts of 5 words each?

Prompt Engineering / GenAI

text = 'Machine learning helps computers learn from data and improve over time'
words = text.split()
chunks = [' '.join(words[i:i+5]) for i in range(0, len(words), 5)]
print(chunks)

A ['Machine learning helps computers', 'learn from data and improve', 'over time']

B ['Machine learning helps computers learn', 'from data and improve over', 'time']

C ['Machine learning helps computers learn from', 'data and improve over time']

D ['Machine learning helps', 'computers learn from data', 'and improve over time']

💡 Hint

Look at how the range and slicing work with step 5.

The code splits the text into words, then groups every 5 words into a chunk. The first chunk has words 0-4, second 5-9, and the last chunk has the remaining word.

❓ Model Choice

advanced

Choosing chunk size for a transformer model

You want to chunk a large document for a transformer model with a maximum input length of 512 tokens. Which chunk size is best to avoid losing context and stay within limits?

A Chunk size of 50 tokens to keep chunks very small.

B Chunk size of 512 tokens to exactly match the model's max input length.

C Chunk size of 1000 tokens to reduce the number of chunks.

D Chunk size of 256 tokens to allow some overlap and context between chunks.

💡 Hint

Think about balancing chunk size and context overlap.

Using chunk size equal to max input length leaves no room for overlap, which can cause loss of context between chunks. Smaller chunks with overlap help preserve context while staying within limits.

❓ Metrics

advanced

Evaluating chunking impact on model accuracy

After chunking text differently, you get these model accuracies on a classification task: - Chunk size 128 tokens: 85% - Chunk size 256 tokens: 88% - Chunk size 512 tokens: 86% Which chunk size likely balances context and input size best?

A 128 tokens, because smaller chunks always improve accuracy.

B 512 tokens, because larger chunks contain more information.

C 256 tokens, because it gives the highest accuracy by balancing chunk size and context.

D All chunk sizes perform the same, so chunking does not matter.

💡 Hint

Look at which chunk size yields the best accuracy.

256 tokens chunk size yields the highest accuracy, indicating it balances enough context and manageable input size for the model.

🔧 Debug

expert

Debugging chunk overlap code

What error or output does this code produce? text = 'AI models need context to understand text better' words = text.split() overlap = 2 chunk_size = 5 chunks = [] for i in range(0, len(words), chunk_size - overlap): chunk = words[i:i+chunk_size] chunks.append(' '.join(chunk)) print(chunks)

A ['AI models need context to', 'context to understand text better', 'text better']

B ['AI models need context to', 'need context to understand', 'to understand text better']

C ['AI models need context to', 'models need context to understand', 'context to understand text better']

D IndexError because the loop steps cause out-of-range slicing.

💡 Hint

Check how the loop increments and slicing work with overlap.

The loop moves by chunk_size - overlap (3), so chunks overlap by 2 words. The slices produce overlapping chunks as in option A.