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.

• ✗

  NER that works across spans (paragraphs) of text rather than single sentences

  Why it's wrong here

  This description confuses the text span as a unit of analysis with the broader context of paragraphs. Span-based NER identifies candidate entity phrases within a sentence and classifies each one, whereas paragraph-level processing is about leveraging surrounding context to improve predictions. The key mechanism is extracting and classifying local text spans, not aggregating information across multiple sentences or paragraphs.

• ✓

  Identifying text spans as candidate entities and classifying each span — handling overlapping and nested entities

  Why this is correct

  Span-based NER works by enumerating all possible text intervals in a sequence and classifying each candidate span as a named entity type or non-entity. Because every span is scored independently, overlapping and nested entities—such as 'University of Washington' inside a larger organization—are handled naturally, whereas token-level sequence labeling typically assumes flat, non-overlapping entities. This direct span classification is the defining characteristic of the approach.

• ✗

  A technique that spans multiple languages to recognise entities in multilingual text

  Why it's wrong here

  This option describes cross-lingual NER, which addresses how well a model transfers entity recognition across different languages, but that is orthogonal to the architectural strategy of span-based NER. Span-based NER is concerned with how entities are extracted and classified from text—by enumerating spans and labeling them—not with the number of languages supported. A span-based model can be monolingual or multilingual; multilingual capability is a separate design consideration.

• ✗

  NER that spans multiple documents to track entities across a corpus

  Why it's wrong here

  Tracking the same entity across many documents requires coreference resolution and entity linking to decide whether mentions in different sources refer to the same real-world object. Span-based NER, by contrast, operates locally within a single document: it extracts candidate text spans and assigns entity types, such as person, location, or organization. It does not attempt to connect those entities across a broader corpus, so this definition conflates span-based extraction with cross-document coreference.

Go deeper

Related to this question