This study is a semantic and pragmatic analysis of the Norwegian right-detached particle så, which occurs exclusively with sentence fragments. The framework of description is relevance theory, a cognitively based theory of communication whose objective is to account for how we are able to understand utterances and to make ourselves understood, in spite of the fact that the linguistic code that we use vastly underdetermines what we mean, and even what we say in the strict sense (truth conditions). It is argued that så encodes a procedure for the addressee to follow in his inferential processing of the linguistic signal. The lexical entry for så contains an instruction to the addressee to contextually activate one or more positive propositions which resemble the one asserted by the utterance of the så-fragment, and to arrange them mentally on a scale whose lower bound is represented by the så-fragment proposition. The speaker implicitly communicates her lack of commitment to the more highly ranked propositions on the scale. The principle behind the scalar ranking is argued to be highly context-dependent.

The focus of my analysis is the so-called existential construction. The languages examined are English, Swedish and Icelandic. The present article assumes the perspective of Ronald W. Langacker's cognitive grammar as the theoretical background. First of all, the assumption is that the unstressed, initial pronoun there, or its Scandinavian equivalents, are semantically definable as abstract-setting subjects of their respective sentences, with, possibly, the exception of Icelandic það. Secondly, the conceptualization of the existential scenes in the three languages is a dynamic process in each case. The dynamicity of the semantics of existential scenes is the result of assuming two planes, the actual and a virtual one, and establishing correspondences between them. The actual plane reflects our direct apprehension of reality. A virtual plane consists in the dynamic re-assignment of roles to the actual elements introduced by means of the virtual abstract-setting subject.

Structural ambiguity, particularly attachment of prepositional phrases, is a serious type of global ambiguity in Natural Language. The disambiguation becomes crucial when a syntactic analyzer must make the correct decision among at least two equally grammatical parse-trees for the same sentence. This paper attempts to find answers to the problem of how attachment ambiguity can be resolved by utilizing Machine Learning (ML) techniques. ML is founded on the assumption that the performance in cognitive tasks is based on the similarity of new situations (testing) to stored representations of earlier experiences (training). Therefore, a large amount of training data is an important prerequisite for providing a solution to the problem. A combination of unsupervised and restricted supervised acquisition of such data will be reported. Training is performed both on a subset of the content of the Gothenburg Lexical Database (GLDB), and on instances of large corpora annotated with coarse-grained semantic information. Testing is performed on corpora instances using a range of different algorithms and metrics. The application language is written Swedish.