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.

• ✗

  Performing sentiment analysis only on the largest datasets to maximise accuracy

  Why it's wrong here

  Sentiment analysis is appropriate for datasets of any size, and accuracy depends on model training and data relevance—not on processing only the largest datasets. The concept of 'at scale' describes a production system's ability to cost-effectively process very large volumes, such as streaming or batch workloads, rather than a requirement to exclude smaller datasets. Applying sentiment only to huge corpora would ignore the value of immediate feedback from small samples.

• ✓

  Efficiently processing large volumes of text for sentiment insights using batch APIs and multilingual support

  Why this is correct

  At-scale sentiment in Azure AI Language uses asynchronous batch APIs to process massive datasets—such as millions of reviews or support tickets—efficiently, with built-in multilingual support for global content. This throughput-focused capability enables organizations to derive actionable sentiment trends and aggregate insights without per-call latency constraints, distinguishing it from improvements in model accuracy or post-processing score calibration.

• ✗

  Scaling the sentiment score range to match industry-standard rating systems

  Why it's wrong here

  Scaling sentiment output to match industry-standard rating systems (e.g., 1–5 stars) is a post-processing normalization step, not what the phrase 'sentiment at scale' means. At-scale sentiment refers to the computational ability to run sentiment analysis on very large text volumes efficiently, typically through batch APIs and parallel processing, without altering the score range itself. Score scaling is about output formatting, while at-scale is about input volume and throughput.

• ✗

  Using larger, more powerful ML models to improve sentiment accuracy on difficult text

  Why it's wrong here

  This option confuses model capability with processing scale: using larger, more powerful ML models may improve sentiment accuracy on ambiguous or domain-specific text, but 'at scale' in AI-900 refers to handling high-volume production workloads through batch APIs and distributed infrastructure. Model size is a quality lever, not a throughput lever, and sentiment-at-scale is defined by the efficient processing of large volumes, not by model complexity.

Go deeper

Related to this question