5.1. NLP Workload Scenarios

💡 First Principle: NLP capabilities analyze different aspects of the same text. Key phrase extraction finds main topics. Entity recognition finds named things (people, places, dates). Sentiment analysis finds emotional tone. Each extracts different information from identical input.

What breaks without this distinction: You'll see scenarios like "analyze customer feedback to identify mentioned products and locations" and need to choose between key phrases, entities, and sentiment. Without understanding what each extracts, you'll guess. But "products and locations" are named entities—that's entity recognition. Key phrases find topics; sentiment finds emotion; entities find specific named things.

Imagine these capabilities like different highlighter colors on a document. Yellow highlights main topics (key phrases). Blue highlights names of people, places, and organizations (entities). Pink highlights emotional words (sentiment). For instance, consider a social media post: "Microsoft announced Azure AI updates in Seattle, and customers love it!" Key phrases: "Azure AI updates". Entities: "Microsoft" (org), "Seattle" (location). Sentiment: positive. Each reveals different information from the same text. What information does your scenario need—topics, names, or feelings? Answer that question and you've identified the capability.

Natural Language Processing enables machines to understand, interpret, and generate human language. Let's examine each capability in detail.