,,Linguistische Bedeutung und Bayes'sche Modellierung"

01.04.2024 bis 31.03.2027

Linguistic analysis is a crucial factor in understanding human communication and its underlying meaning. For more than half a century, linguists have relied on logical and formal methods for detailed analysis. These methods have proven very successful in explaining how the literal meaning of sentences and text segments is composed of the meaning of their parts. However, logical methods reach their limits when it comes to context-dependent meaning as it emerges in natural communicative settings with various sources of uncertainly and contextual expectations about communicative goals. These limitations turned out to be increasingly critical with the advent of experimental pragmatics, which provides detailed detain communicated meaning and reveals the probabilistic nature of language use and interpretation. our goal is to overcome these limitations following the lead of recent attempts to supplement logical analysis by game theoretic methods and Bayesian statistical modeling. We want to be able to simulate the speakers choose between linguistic alternatives in ecologically valid scenarios and predict how semantic meaning is modified by the interaction between interlocutors, relevance defined by questions under discussion, and cognitive constraints such as perceptual imprecision and uncertainty about objectives and the state affairs. Ecological validity requires the ability to cope with highly complex models. To reach our goals and upscale models, statistical problems have to be solved that cannot be overcome by linguists alone. Problems range from meaningful criteria of model evaluation and comparisons methods for fitting high dimensional models, the elimination of purely framework related restrictions, to issues like the integration of Bayesian cognitive models in game theoretic frameworks. We expect a breakthrough from the direct collaboration between linguists and mathematicians.