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

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the purpose of hierarchical chunking

What is the main advantage of using hierarchical chunking in machine learning models?

A It increases the speed of training by using only the first layer of data representation.

B It allows models to process data at multiple levels of abstraction, improving understanding of complex structures.

C It reduces the size of the dataset by removing irrelevant features before training.

D It guarantees 100% accuracy by memorizing all training examples.

💡 Hint

Think about how breaking data into parts helps models learn better.

Hierarchical chunking helps models learn by breaking data into meaningful parts at different levels, allowing better capture of complex patterns.

❓ Predict Output

intermediate

Output of hierarchical chunking on a sequence

Given the following Python code that simulates hierarchical chunking on a sequence of tokens, what is the output?

Prompt Engineering / GenAI

def hierarchical_chunking(tokens):
    level1 = [tokens[i:i+2] for i in range(0, len(tokens), 2)]
    level2 = [level1[i:i+2] for i in range(0, len(level1), 2)]
    return level2

sequence = ['I', 'love', 'machine', 'learning', 'and', 'AI']
result = hierarchical_chunking(sequence)
print(result)

A [[['I', 'love'], ['machine', 'learning']], [['and', 'AI']]]

B [[['I', 'love', 'machine'], ['learning', 'and']], [['AI']]]

C [['I', 'love'], ['machine', 'learning'], ['and', 'AI']]

D [['I', 'love', 'machine', 'learning'], ['and', 'AI']]

💡 Hint

Look at how the code groups tokens first in pairs, then groups those pairs again.

The code first chunks tokens into pairs (level1), then groups those pairs into pairs again (level2), resulting in nested lists.

❓ Model Choice

advanced

Choosing a model architecture for hierarchical chunking

Which model architecture is best suited to effectively learn hierarchical chunking representations in natural language processing?

A Hierarchical Transformers that process data at multiple levels of granularity

B Convolutional Neural Networks (CNNs) with fixed-size filters only

C Recurrent Neural Networks (RNNs) with simple sequential processing

D Linear regression models without hidden layers

💡 Hint

Consider models that can capture relationships at different scales.

Hierarchical Transformers can process input at multiple levels, making them ideal for hierarchical chunking tasks.

❓ Hyperparameter

advanced

Hyperparameter affecting hierarchical chunking depth

In a hierarchical chunking model, which hyperparameter directly controls the depth of chunking layers?

A Batch size

B Learning rate

C Number of hierarchical layers or levels

D Dropout rate

💡 Hint

Think about what controls how many nested chunks the model creates.

The number of hierarchical layers determines how many levels of chunking the model performs.

❓ Metrics

expert

Evaluating hierarchical chunking model performance

You trained a hierarchical chunking model for text segmentation. Which metric best measures how well the model identifies correct chunk boundaries?

A Training loss value after the first epoch

B Mean squared error of token embeddings

C Perplexity of the language model on the training data

D Accuracy of predicted chunk boundaries compared to true boundaries

💡 Hint

Focus on how well the model finds the right places to split the text.

Accuracy of predicted chunk boundaries directly measures how well the model segments text into correct chunks.