LT World

You are here: Home kb Information & Knowledge Technologies Natural Language Generation

Natural Language Generation

Building Natural Language Generation Systems.
Reiter, Ehud and Dale, Robert.
Cambridge University Press. Cambridge UK. 2000.

Planning Text for Advisory Dialogues: Capturing Intentional and Rhetorical Information.
Johanna Moore and Cecile Paris.
Computational Linguistics. 19 (4). 1993. 651-694.

Text Generation: Using Discourse Strategies and Focus Constraints to Generate Natural Language Text.
Kathleen R. McKeown.
Cambridge University Press. Cambridge. 1985.

Generierung natürlichsprachlicher Texte.
Stephan Busemann.
Günther Görz, Claus-Rainer Rollinger, Josef Schneeberger. Künstliche Intelligenz. Oldenburg Verlag, München, Wien. 2000. 783-814.

  • Natural Language Generation Group at the Department of Computing Science, University of Aberdeen (NLG Group Aberdeen)
  • Columbia Natural Language Processing Group
  • CoGenTex, Inc.
  • German Research Center for Artificial Intelligence (DFKI)
  • Institute for Communicating and Collaborative Systems (ICCS)
  • University of Aberdeen
  • Natural Language Technology Group (NLTG)
  • Special Interest Group on Generation (SIGGEN)
  • Natural Language Generation in Saarbrücken (NLG)

  • Roger Evans
  • Robert Dale
  • Chris Mellish
  • Helmut Horacek
  • Donia Scott
  • Mariët Theune
  • Matthew Stone
  • Kathleen R. McKeown
  • Anja Belz
  • David McDonald
  • Tilman Becker
  • Emiel Krahmer
  • Claire Gardent
  • John A. Bateman
  • Michael Elhadad
  • Johanna D. Moore
  • Kees Van Deemter
  • Cecile L. Paris
  • Manfred Stede
  • Michael White
  • Ehud Reiter
  • Stephan Busemann
  • Charles Callaway

  • RealPro
  • ProjectReporter
  • TG/2

Natural Language Generation (NLG) is concerned with turning some usually non-linguistic representation of information and intended effect into fluent text preserving both meaning and intention.

NLG systems often identify the content to be verbalized. They structure the document into interrelated sentence-sized chunks, choose appropriate words, aggregate and elide information to ensure fluency, create contextually appropriate referring expressions, such as pronouns, and follow grammatical constraints of the chosen language. All this is achieved using knowledge about the world and the domain of dicsourse, about communication and about languages.

NLG components are used for e.g. automatic report generation, document authoring, dialogue, concept-to-speech, multi-modal and machine translation systems.

Evaluating the correctness and the appropriateness of generated text is a research theme on its own since there is usually no single correct solution. One important way to tackle the problem consists in creating reference corpora and performing shared evaluation tasks, e.g. on generating referring expressions. However, this is not intended to replace less formal evaluation strategies such as human assessments.