• Digital Humanities and Cultural Heritage
• Publications close
• Research data close
• Research software close
• Other research products close
• Article close
• EU close
• IT close
• HELDA - Digital Repository of the U... close

In our submission to the SIGMORPHON 2022 Shared Task on Morpheme Segmentation, we study whether an unsupervised morphological segmentation method, Morfessor, can help in a supervised setting. Previous research has shown the effectiveness of the approach in semisupervised settings with small amounts of labeled data. The current tasks vary in data size: the amount of word-level annotated training data is much larger, but the amount of sentencelevel annotated training data remains small. Our approach is to pre-segment the input data for a neural sequence-to-sequence model with the unsupervised method. As the unsupervised method can be trained with raw text data, we use Wikipedia to increase the amount of training data. In addition, we train multilingual models for the sentence-level task. The results for the Morfessor-enriched features are mixed, showing benefit for all three sentencelevel tasks but only some of the word-level tasks. The multilingual training yields considerable improvements over the monolingual sentence-level models, but it negates the effect of the enriched features. Peer reviewed

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

Open access airborne laser scanning (ALS) data have been available in Finland for over a decade and have been actively applied by the Finnish archaeologists in that time. The low resolution of this laser scanning 2008-2019 dataset (0.5 points/m(2)), however, has hindered its usability for archaeological prospection. In the summer of 2020, the situation changed markedly, when the Finnish National Land Survey started a new countrywide ALS survey with a higher resolution of 5 points/m(2). In this paper we present the first results of applying this newly available ALS material for archaeological studies. Finnish LIDARK consortium has initiated the development of semi-automated approaches for visualizing, detecting, and analyzing archaeological features with this new dataset. Our first case studies are situated in the Alpine tundra environment of Sapmi in northern Finland, and the assessed archaeological features range from prehistoric sites to indigenous Sami reindeer herding features and Second Word War-era German military structures. Already the initial analyses of the new ALS-5p data show their huge potential for locating, mapping, and assessing archaeological material. These results also suggest an imminent burst in the number of known archaeological sites, especially in the poorly accessible and little studied northern wilderness areas, when more data become available. Peer reviewed

Surface wave (SW) methods offer promising options for an effective and sustainable development of seismic exploration, but they still remain under-exploited in hard rock sites. We present a successful application of active and passive surface wave tomography for the characterization of the southern continuation of the Siilinjarvi phosphate deposit (Finland). A semi-automatic workflow for the extraction of the path-average dispersion curves (DCs) from ambient seismic noise data is proposed, including identification of time windows with strong coherent SW signal, azimuth analysis and two-station method for DC picking. DCs retrieved from passive data are compared with active SW tomography results recently obtained at the site. Passive data are found to carry information at longer wavelengths, thus extending the investigation depth. Active and passive DCs are consequently inverted together to retrieve a deep pseudo-3D shearwave velocity model for the site, with improved resolution. The southern continuation of the mineralization, its contacts with the host rocks and different sets of cross-cutting diabase dikes are well imaged in the final velocity model. The seismic results are compared with the latest available geological models to both validate the proposed workflow and improve the interpretation of the geometry and extent of the mineralization. Important large-scale geological boundaries and structural discontinuities are recognized from the results, demonstrating the effectiveness and advantages of the methods for mineral exploration perspectives. Peer reviewed

Adjectives like pretty, beautiful and gorgeous describe positive properties of the nouns they modify but with different intensity. These differences are important for natural language understanding and reasoning. We propose a novel BERT-based approach to intensity detection for scalar adjectives. We model intensity by vectors directly derived from contextualised representations and show they can successfully rank scalar adjectives. We evaluate our models both intrinsically, on gold standard datasets, and on an Indirect Question Answering task. Our results demonstrate that BERT encodes rich knowledge about the semantics of scalar adjectives, and is able to provide better quality intensity rankings than static embeddings and previous models with access to dedicated resources. Comment: EMNLP 2020

There are several approaches for improving neural machine translation for low-resource languages: monolingual data can be exploited via pretraining or data augmentation; parallel corpora on related language pairs can be used via parameter sharing or transfer learning in multilingual models; subword segmentation and regularization techniques can be applied to ensure high coverage of the vocabulary. We review these approaches in the context of an asymmetric-resource one-to-many translation task, in which the pair of target languages are related, with one being a very low-resource and the other a higher-resource language. We test various methods on three artificially restricted translation tasks—English to Estonian (low-resource) and Finnish (high-resource), English to Slovak and Czech, English to Danish and Swedish—and one real-world task, Norwegian to North Sámi and Finnish. The experiments show positive effects especially for scheduled multi-task learning, denoising autoencoder, and subword sampling. There are several approaches for improving neural machine translation for low-resource languages: monolingual data can be exploited via pretraining or data augmentation; parallel corpora on related language pairs can be used via parameter sharing or transfer learning in multilingual models; subword segmentation and regularization techniques can be applied to ensure high coverage of the vocabulary. We review these approaches in the context of an asymmetric-resource one-to-many translation task, in which the pair of target languages are related, with one being a very low-resource and the other a higher-resource language. We test various methods on three artificially restricted translation tasks-English to Estonian (low-resource) and Finnish (high-resource), English to Slovak and Czech, English to Danish and Swedish-and one real-world task, Norwegian to North Sami and Finnish. The experiments show positive effects especially for scheduled multi-task learning, denoising autoencoder, and subword sampling. Peer reviewed