The University of Aberdeen

Course Title: Natural Language Generation

Course Organiser: Advaith Siddharthan


It is no longer only in Science Fiction that computers are able to explain data and other information to us in natural languages like English and French. For instance, weather reports on the UK Met Office website are generated using programs written here at Aberdeen, through a spinoff company Data2Text that has now been acquired by Arria NLG plc. On the other hand, there are still large gaps between computer and human performance on many natural language generation tasks. This course will review basic techniques for natural language generation, such as content selection, microplanning and surface realisation, and cover a range of applications ranging from weather forcasts to writing reports from scuba-diving logs. The course aims to:

  1. Explain the basics of language, such as vocabulary, grammar and discourse
  2. Explain why natural language generation is hard
  3. Give examples of state-of-the-art natural language generation systems
  4. Present useful general concepts and algorithms for building natural language generation systems

The course lasts for 12 weeks and is worth 15 credits. The course organiser is Advaith Siddharthan. Lecturers are Kees van Deemter, Advaith Siddharthan, as well as guest lecturers from Data2Text (now Arria NLG), an Aberdeen University startup company developing data communication solutions based on Natural Language Generation.

Timetable

Lectures Practicals Practicals start in the second week of teaching.