This paper explores the use of linguistic strategies, specifically discourse markers like 'but', to express contrasts between expectations and reality when faced with unexpected events. The study concentrates on Denial of Expectation (DofE), the
most powerful form of contrast, which arises when the expected value based on background assumptions is not met. The main focus of this paper is to model DofE as a weighted homogeneous relationship between object properties. The aim is to predict DofE for numerical properties in specific contexts. I aim to address a gap in previous models by considering the role of context. This is achieved by analyzing contrastive sentences from German car and motorcycle reviews. The research presents the concept of expectation intervals for scalar properties. These intervals align with expectations and exceeding them triggers a potential contrast. The study incorporates causality, expected behavior, and a shift function in selecting contrastive pairs, transforming the conditions into an algorithm.
Keywords: contrast; computational and cognitive modeling; discourse analysis