The talk will introduce the major issues and techniques in the field of natural language generation (NLG). First, the benefits of NLG in contrast to simpler techniques such as template filling will be discussed. Then, the different levels of generation will be presented: discourse and context planning, sentence planning, lexical choice, and linearization into actual text. Following the historical development of the field, the first part of the talk will use examples from knowledge-based generation. I will mention some of the major systems and approaches, including software tools for performing the different generation steps, and discuss how generation technology has been used in applications such as medical information systems, multimedia co-ordination, speech synthesis, and summarization. The second part of the talk will discuss more recent achievements using empirical, data-driven methods. Techniques that use text corpora to learn how to perform different generation stages will be presented, with examples from lexical choice, sentence ordering, and context plan learning. The talk assumes minimal exposure to computational linguistics and no previous background in natural language generation.