• Digital Humanities and Cultural Heritage
• Publications close
• FI close
• NARCIS close

In European and many African, Middle Eastern and southern Asian populations, lactase persistence (LP) is the most strongly selected monogenic trait to have evolved over the past 10,000 years1. Although the selection of LP and the consumption of prehistoric milk must be linked, considerable uncertainty remains concerning their spatiotemporal configuration and specific interactions2,3. Here we provide detailed distributions of milk exploitation across Europe over the past 9,000 years using around 7,000 pottery fat residues from more than 550 archaeological sites. European milk use was widespread from the Neolithic period onwards but varied spatially and temporally in intensity. Notably, LP selection varying with levels of prehistoric milk exploitation is no better at explaining LP allele frequency trajectories than uniform selection since the Neolithic period. In the UK Biobank4,5 cohort of 500,000 contemporary Europeans, LP genotype was only weakly associated with milk consumption and did not show consistent associations with improved fitness or health indicators. This suggests that other reasons for the beneficial effects of LP should be considered for its rapid frequency increase. We propose that lactase non-persistent individuals consumed milk when it became available but, under conditions of famine and/or increased pathogen exposure, this was disadvantageous, driving LP selection in prehistoric Europe. Comparison of model likelihoods indicates that population fluctuations, settlement density and wild animal exploitation—proxies for these drivers—provide better explanations of LP selection than the extent of milk exploitation. These findings offer new perspectives on prehistoric milk exploitation and LP evolution. 336-345 200

Semantic representations have long been argued as potentially useful for enforcing meaning preservation and improving generalization performance of machine translation methods. In this work, we are the first to incorporate information about predicate-argument structure of source sentences (namely, semantic-role representations) into neural machine translation. We use Graph Convolutional Networks (GCNs) to inject a semantic bias into sentence encoders and achieve improvements in BLEU scores over the linguistic-agnostic and syntax-aware versions on the English--German language pair.

Retrospective data harmonization is an open issue in healthcare due to the emerging need to interlink data from multiple clinical centers with the absence of standardized data collection protocols. In this work, we present an automated data harmonization workflow which utilizes lexical and semantic analysis based on word embeddings and relational modeling to detect terminologies with common lexical and conceptual basis. The method is built on top of a knowledge base to enable the interlinking of heterogeneous cross-domain data. A case study is applied in two clinical domains, namely the cardiovascular disease (CVD) and the mental disorders, where the proposed method yielded matched terminologies with 85% precision in less execution time than the application of lexical analysis and manual mapping which yielded 10% less precision.

Medical research would benefit from automatic methods that support eligibility evaluation for patient enrollment in clinical trials and design of eligibility criteria. In this study we addressed the problem of formalizing eligibility criteria. By analyzing a large set of breast cancer clinical trials we derived a set of patterns, that capture typical structure of conditions, pertaining to syntax and semantics. We qualitatively analyzed their expressivity and evaluated coverage using regular expressions, running experiments on a few thousands of clinical trials also related to other diseases. Based on an early evaluation we conclude that derived patterns cover the language of eligibility criteria to a large extent and may serve as a semi-formal representation. We expect that extending the presented method for pattern recognition with recognition of ontology concepts will facilitate generating computable queries and automated reasoning for various applications.

Northeastern Siberia has been inhabited by humans for more than 40,000 years but its deep population history remains poorly understood. Here we investigate the late Pleistocene population history of northeastern Siberia through analyses of 34 newly recovered ancient genomes that date to between 31,000 and 600 years ago. We document complex population dynamics during this period, including at least three major migration events: an initial peopling by a previously unknown Palaeolithic population of ‘Ancient North Siberians’ who are distantly related to early West Eurasian hunter-gatherers; the arrival of East Asian-related peoples, which gave rise to ‘Ancient Palaeo-Siberians’ who are closely related to contemporary communities from far-northeastern Siberia (such as the Koryaks), as well as Native Americans; and a Holocene migration of other East Asian-related peoples, who we name ‘Neo-Siberians’, and from whom many contemporary Siberians are descended. Each of these population expansions largely replaced the earlier inhabitants, and ultimately generated the mosaic genetic make-up of contemporary peoples who inhabit a vast area across northern Eurasia and the Americas. Analyses of 34 ancient genomes from northeastern Siberia, dating to between 31,000 and 600 years ago, reveal at least three major migration events in the late Pleistocene population history of the region.

We present the RUG-SU team's submission at the Native Language Identification Shared Task 2017. We combine several approaches into an ensemble, based on spelling error features, a simple neural network using word representations, a deep residual network using word and character features, and a system based on a recurrent neural network. Our best system is an ensemble of neural networks, reaching an F1 score of 0.8323. Although our system is not the highest ranking one, we do outperform the baseline by far.