Natural Language Generation (NLG)

The use of artificial intelligence programming to produce written or spoken narratives from a dataset.

How it works

NLG systems transform structured data or internal representations into coherent natural language text. Traditional NLG pipelines had discrete stages: content determination (what to say), document planning (how to organise it), microplanning (choosing words and sentence structures), and surface realisation (producing grammatical text). Modern LLM-based NLG replaces most of these stages with a single end-to-end model that generates text autoregressively from a prompt and structured data input.

Why it matters

NLG enables computers to communicate with humans in their own language rather than forcing humans to interpret raw data or structured reports. Applications include automated financial reports, personalised weather summaries, chatbot responses, and code documentation generation. As LLM quality has improved, NLG has become a commodity capability that dramatically reduces the cost of producing large volumes of personalised, high-quality written content.