Aim
Based on results from computational historical linguistics and statistical typology, CrossLingferences goal is to establish a probabilistic framework to analyze change and diversification processes in language. This enables the investigation of issues located at the interface of historical linguistics and typology. On the one hand, existing methods and algorithms should be adapted and on the other hand novel approaches should be developed to answer these questions. Special attention will be paid to hierarchical Bayesian methods.
Work Package 1: Lexical Reconstruction
Based on the resources of the NorthEuralex database lexical reconstruction should be done using existing algorithms. In a second step, novel algorithms should be developed which are able to generalize across different language families.
Work Package 2: Hierarchical phylogenetic Markov chain and multinomial models
The aim of this work package is the development of algorithms which are tailored to the specific needs for the phylogenetic comparative methods. Especially, the capability to analyze multiple phylogenies over disjoint taxon sets and the availability of several birth-death processes is necessary.
Work Package 3: Language contact and spatial random effects
Typological variation is known to be heavily influenced by language contact. In this work package language contact should be analyzed using auto-logistic autoregressive models.
Work Package 4: Approximate Bayesian Computation and agent-based simulations
While the models which will be developed in work packages 1-3 are influenced by traditional MCMC variants, the aim of this work package is to explore the use cases of Approximate Bayesian Computation. Especially models for lexical reconstruction and language contact should be developed which are intractable using classical methods.
Work Package 5: Causal Inference
The goal of this package is to identify potential causal links between typological variables. The inference is based on results of work packages 1 and 3.
Work Package 6: Software library
To realize the potential impact of the CrossLingference, the algorithms and methods developed in this project should be made available as a software library to the scientific community.