A stochastic local search approach to language tree reconstruction

In this paper we introduce a novel stochastic local search algorithm to reconstruct phylogenetic trees. We focus in particular on the reconstruction of language trees based on the comparison of the Swadesh lists of the recently compiled ASJP database. Starting from a generic tree configuration, our scheme stochastically explores the space of possible trees driven by the minimization of a pseudo-functional quantifying the violations of additivity of the distance matrix. As a consequence the resulting tree can be annotated with the values of the violations on each internal branch. The values of the deviations are strongly correlated with the stability of the internal edges; they are measured with a novel bootstrap procedure and displayed on the tree as an additional annotation. As a case study we considered the reconstruction of the Indo-European language tree. The results are quite encouraging, highlighting a potential new avenue to investigate the role of the deviations from additivity and check the reliability and consistency of the reconstructed trees.

Published online: 26 September 2012
https://doi.org/10.1075/bct.46.09tri