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.

• ✗

  Key Phrase Extraction

  Why it's wrong here

  Key Phrase Extraction identifies the most prominent words and phrases in a document based on statistical importance, but it outputs untyped, surface-level strings rather than structured medical entities. It has no clinical model, no entity type labels, and no relation or negation detection, so it cannot reliably distinguish a medication from a symptom or capture dosage values. This makes it unsuitable for converting thousands of clinical narratives into a structured dataset for research.

  When this WOULD be correct

  A company wants to automatically extract the most important points from customer feedback surveys to identify common themes. Key Phrase Extraction would be correct because it extracts general key phrases without needing domain-specific entities.

• ✗

  Named Entity Recognition (NER)

  Why it's wrong here

  Standard Named Entity Recognition (NER) in Azure AI Language is a generic model that classifies entities into broad categories like person, location, organization, and date/time. It is not trained on medical ontologies, so it does not recognize clinical concepts such as drug names or disease codes with clinical accuracy. Using it for medical research would require extensive downstream mapping and would miss the domain-specific relations and assertions that clinical text demands.

  When this WOULD be correct

  A company needs to extract general entities like person names, locations, and dates from legal contracts. They require a prebuilt Azure AI Language feature that identifies common entity types without custom training.

• ✓

  Text Analytics for Health

  Why this is correct

  Text Analytics for Health is a purpose-built, pre-trained Azure AI Language service specifically designed for clinical text. It extracts medical entities such as diseases, symptoms, medications, and dosages while also capturing relations, negation, and assertion context. It links extracted concepts to standard medical ontologies like UMLS and ICD-10-CM, making it immediately applicable to processing thousands of clinical notes without custom model training.

• ✗

  Custom Text Classification

  Why it's wrong here

  Custom Text Classification is a supervised machine learning capability where you provide labeled document-level categories and train a model to assign those labels to entire documents or sentences. It does not extract entity-level data such as a specific disease or medication from within the text, and it would require building a large labeled clinical dataset and training a bespoke model. This is neither prebuilt nor geared toward medical entity extraction, so it cannot be deployed immediately to process clinical documents at scale.

  When this WOULD be correct

  A research organization needs to classify clinical trial documents into custom categories (e.g., 'efficacy', 'safety') based on their own labeling scheme, and they have a labeled dataset to train a model. In that scenario, Custom Text Classification 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