Microsoft Certified Azure AI Fundamentals · Free Practice Question Medium

Question 1

A company is developing a chatbot to categorize customer feedback into predefined sentiment categories (e.g., "positive," "neutral," "negative"). The model struggles to group phrases like "extremely satisfied" and "delighted" into the same category.

What should be prioritized to improve the model's understanding of semantic similarity between phrases?

  • A

    Use vector embeddings to represent the text.

  • B

    Convert all text to lowercase before processing.

  • C

    Add more explicit instructions in the prompt.

  • D

    Increase the number of tokens processed per input.

Reveal correct answer

Correct answer: A

Explanation

  1. What Vector Embeddings Do:
    Vector embeddings convert words or phrases into numerical representations in a continuous vector space. In this space, semantically similar phrases are mapped to vectors that are close to one another. This means that phrases like "extremely satisfied" and "delighted" will have similar vector representations, helping the model to correctly identify them as conveying similar sentiment.

    • Example: By using embeddings, the model can learn that these phrases, despite being different in wording, share a similar semantic meaning, thereby grouping them into the same sentiment category.

  2. Why Vector Embeddings Are the Best Choice:

    • Enhanced Semantic Understanding: Vector embeddings capture contextual relationships between words, which is crucial for understanding nuanced expressions in customer feedback.

    • Model Improvement: Incorporating embeddings into the sentiment analysis pipeline enables the model to better distinguish subtle variations in sentiment, leading to more accurate categorization.

    • Standard Practice in NLP: Utilizing embeddings is a well-established method in natural language processing and is widely recommended for tasks involving semantic similarity and sentiment analysis.

  3. Why the Other Options Are Not Suitable:

    • Increase the Number of Tokens Processed per Input:

      • Issue: While processing more tokens might allow the model to consider more context, it does not inherently improve the model’s understanding of semantic relationships between words.

    • Convert All Text to Lowercase Before Processing:

      • Issue: Lowercasing standardizes text format but does not address the deeper semantic meaning or similarity between different phrases.

    • Add More Explicit Instructions in the Prompt:

      • Issue: Although providing explicit instructions can help guide the model, it does not fundamentally change the representation of the text. The core issue of capturing semantic similarity is best addressed by embedding the text into a meaningful vector space.

Official Documentation References:

A.

Vector embeddings represent words/phrases in a multidimensional space where similar meanings are mathematically closer. This helps the model recognize semantic relationships (e.g., "delighted" and "extremely satisfied" would have similar embeddings).

B.

Lowercase conversion standardizes text but doesn’t resolve semantic gaps.

C.

Instructions guide tasks but don’t directly address semantic grouping.

D.

Token count affects input size but doesn’t inherently improve semantic understanding.

Discussion

Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.

You must be logged in to post a comment.

Preparing For

Your Certification?

255+ certifications
Detailed explanations
Free PDF samples

Has All The Questions You Need