Direct Probabilistic Semantics: A Contextualist Formal Semantic Model

This thesis should be viewed in the context of the debate that is currently going on in the field of philosophy of language: the debate on the question of how much influence the context in which a sentence is used has on the meaning of that sentence. Closely related to this is the question of whether a sentence can have a meaning on its own, or if a sentence only has a meaning within the context and circumstances in which it is used. Simplifying a bit, two approaches can be distinguished: that of literalism and of contextualism. Literalism tries to understand the meaning of language by representing it as a formula in a symbolic logic. Word meanings are taken from a lexicon as logical formulas, which are then composed according to the syntactic structure of a sentence to form the meaning of the whole sentence, which is a larger logical formula. Literalists include several mechanisms to introduce contextual information into this process, but according to contextualists these mechanisms are too limited. According to the more extreme contextualists, a sentence can be gotten to mean (almost) anything if it is used in the right context or situation. I find that the arguments the contextualists make are generally convincing, and that a more contextual approach is needed. However one of the main drawbacks of contextualism is that the theories it proposes are not very well developed in a formal sense. This debate takes place mostly in the field of philosophy of language, and the theories proposed by contextualists are also usually formulated in philosophical terms and at a philosophical level of abstraction. On the other hand, literalists have models that are better developed and formalized well enough to program into a computer so that calculations and predictions can be made about them. Contextualist models usually do not reach this level of formal development, and it is therefore no surprise that in the field of formal linguistics the theories promoted by contextualists do not find a large audience. In this thesis I aim to take a stab at this shortcoming. I have developed a model of semantics that is based on the ideas defended by contextualism, and implemented it in a computer program. An important property in which the model differs from many existing theories is that it is a non-truthconditional theory of meaning, and instead implements a more internalist approach. This does not prevent formalization and implementation in a computer, and in fact helps to answer some common philosophical problems. The resulting computer program has the ability to interpret and act upon simple sentences that are about a small scene containing a few geometric objects