AI-900 Practice Question: Describe features of Natural Language Processing workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  A. Text Analytics (prebuilt)

  Why this is correct

  Text Analytics (prebuilt) is the correct choice because it is an out-of-the-box Azure AI Language capability that performs both sentiment analysis and key phrase extraction on raw text without any custom training. The service returns a sentiment score (positive, negative, neutral) per document or sentence, plus a list of key talking points, exactly matching the requirement to analyze chat transcripts. Because it is a prebuilt API, no labeled dataset or model training pipeline is needed.

• ✗

  B. Custom Text Classification

  Why it's wrong here

  Custom Text Classification is incorrect because it requires a labeled dataset and a custom training run to learn to assign predefined categories or classes to documents. Even after training, it only outputs a class label (or multiple labels) per document; it does not provide a sentiment score or extract key phrases. Since the scenario specifies using a prebuilt feature without custom training, a custom model is unnecessary and off-target for the stated needs.

  When this WOULD be correct

  If the question asked for a solution to classify chat transcripts into custom categories (e.g., specific product issue types) that are not predefined by Azure, and the user has labeled data to train a model, then Custom Text Classification would be correct.

• ✗

  C. Conversational Language Understanding

  Why it's wrong here

  Conversational Language Understanding is incorrect because it is designed to identify a user's intent and extract entities from a single utterance within a dialogue, enabling a bot to respond appropriately. It requires custom training on sample utterances and does not provide sentiment analysis or key phrase extraction as part of its core function. The scenario calls for understanding tone and main points in existing transcripts, not for building intents and entities to drive a conversational flow.

  When this WOULD be correct

  A question where the task is to build a custom model to understand user intents and extract entities from customer support conversations, such as identifying whether a user wants to 'reset password' or 'check order status', and extracting the product name. In that case, CLU would be the correct choice.

• ✗

  D. Question Answering

  Why it's wrong here

  Question Answering is incorrect because it is built to derive concise answers from a set of supplied documents (such as FAQs or product manuals) by matching a user's question to a stored QnA pair or an answer span. It does not produce sentiment scores or extract key themes from a conversation. The task requires analyzing chat transcripts for sentiment and key phrases, not retrieving answer spans from a knowledge base.

  When this WOULD be correct

  A company wants to build a bot that can answer customer questions by retrieving answers from a set of FAQ documents or manuals. The prebuilt Question Answering feature would be the correct choice.

Option-by-option analysis

Why each answer is right or wrong

Understanding why wrong answers are wrong — and when they would be correct — is what separates a 750 score from a 900. The AI-900 exam frequently reuses these exact scenarios with slightly different constraints.

Analysis generated from the official AI-900blueprint and verified against question context. The “when correct” sections are what AI assistants cite when candidates ask “what’s the difference between these options?”

Go deeper

Related to this question