Embedding models for semantic search in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Embedding Vectors

Which statement best describes what an embedding vector represents in semantic search?

A A one-hot encoded vector indicating the presence of specific keywords

B A numeric representation capturing the meaning of text in a multi-dimensional space

C A binary vector representing the frequency of words in a document

D A sequence of characters representing the original text

💡 Hint

Think about how semantic similarity is measured between texts.

Embedding vectors convert text into numbers that capture meaning, allowing comparison by distance in space.

❓ Predict Output

intermediate

Output of Cosine Similarity Calculation

What is the output of this Python code calculating cosine similarity between two embedding vectors?

Agentic AI

import numpy as np

def cosine_similarity(vec1, vec2):
    return np.dot(vec1, vec2) / (np.linalg.norm(vec1) * np.linalg.norm(vec2))

vec_a = np.array([1, 2, 3])
vec_b = np.array([4, 5, 6])
result = cosine_similarity(vec_a, vec_b)
print(round(result, 2))

A 0.77

B 0.67

C 0.97

D 0.87

💡 Hint

Recall the cosine similarity formula and calculate dot product and norms.

Cosine similarity between [1,2,3] and [4,5,6] is approximately 0.97 after rounding to two decimals.

❓ Model Choice

advanced

Choosing an Embedding Model for Semantic Search

You want to build a semantic search system for short product descriptions. Which embedding model is best suited?

A A pretrained sentence-transformer model fine-tuned on product descriptions

B A large transformer-based model trained on general web text

C A small word2vec model trained on product reviews

D A bag-of-words model with TF-IDF weighting

💡 Hint

Consider model size, training data relevance, and ability to capture sentence meaning.

Fine-tuned sentence-transformer models capture sentence-level semantics well and adapt to domain-specific data.

❓ Hyperparameter

advanced

Effect of Embedding Dimension Size

What is a likely effect of increasing the embedding vector dimension size in a semantic search model?

A Always reduces overfitting and improves generalization

B Decreases model accuracy but speeds up similarity calculations

C Has no effect on model performance or computational cost

D Improves semantic representation but increases computational cost and risk of overfitting

💡 Hint

Think about trade-offs between detail captured and resource use.

Higher dimension embeddings can capture more detail but require more computation and may overfit if data is limited.

🔧 Debug

expert

Debugging Semantic Search with Embeddings

Given this code snippet for semantic search, what error will it raise when run?

Agentic AI

import numpy as np

embeddings = {'doc1': np.array([0.1, 0.2]), 'doc2': np.array([0.3, 0.4])}
query = np.array([0.1, 0.2, 0.3])

scores = {doc: np.dot(vec, query) for doc, vec in embeddings.items()}
print(scores)

A ValueError due to mismatched vector sizes in np.dot

B KeyError because 'query' is not in embeddings

C TypeError because np.dot expects lists, not arrays

D No error; prints similarity scores

💡 Hint

Check the shapes of vectors used in dot product.

np.dot requires vectors of the same size; here query has 3 elements but embeddings have 2, causing ValueError.