Natural Language Generation (NLG)

Natural Language Generation, or NLG, covers systems that produce written or spoken language. Applications include summarization, translation, dialogue, report generation, data-to-text explanations, and creative assistance. Earlier NLG pipelines often separated content selection, sentence planning, and surface realization, while modern language models can generate complete passages from prompts and context. Quality includes more than grammatical fluency: output should be relevant, factual when required, coherent, appropriately styled, and safe for its audience. Evaluation uses task-specific metrics and human review because many prompts have multiple valid responses. Production systems may also need grounding, templates, controlled vocabulary, citations, and human approval.