A Sentence Uttered Makes a World Appear---Natural Language Interpretation as Abductive Model Generation


What happens when a sentence is interpreted? W.H. Auden, in his sonnet Words, gives the following answer.

A sentence uttered makes a world appear
Where all things happen as it says they do [...]

We think (as others have done) that the poet is right. But if he is, the follow-up scientific question is: How do worlds appear? What is the mechanics of world creation? Our aim is to capture this in a precise logical theory and to base a testable theory of natural language interpretation on the result. Two techniques that were developed in the field of Computational Logic / Artificial Intelligence will be borrowed and adapted: tableau model generation and tableau abduction. Model generation is a formal counterpart to world creation, abduction an ubiquitous form of reasoning that can nicely explain how new information in a sentence connects to old and how presuppositions can be accommodated. The two are merged into a
framework of abductive model generation, which can be used to dynamically represent ongoing discourse.

We ask funding for three postdocs who will work on natural language interpretation as abductive model generation. One will focus on logical questions to do with the idea, a second on linguistic questions, and a third on an application involving the definition of a controlled language. We also ask money for postdoc travel, two conferences, and a website.


Project number


Main applicant

Dr. R.A. Muskens

Affiliated with

Universiteit van Amsterdam, Faculteit der Natuurwetenschappen, Wiskunde en Informatica, Institute for Logic, Language and Computation (ILLC)

Team members

Dr. G. Pulcini


01/12/2019 to 31/10/2024