• Digital Humanities and Cultural Heritage
• 2012-2021close
• Publications close
• Other research products close
• Preprint close
• European Commission close
• Science Foundation Ireland close

Abstract. A unique stratigraphic sequence of fossil leaves of Eotrigonobalanus furcinervis (extinct trees of the beech family, Fagaceae) from central Germany has been used to derive an atmospheric pCO2 record with multiple data points spanning the late middle to late Eocene, two sampling levels which may be earliest Oligocene, and two samples from later in the Oligocene. Using the inverse relationship between the density of stomata and pCO2, we show that pCO2 decreased continuously from the late middle to late Eocene, reaching a relatively stable low value before the end of the Eocene. Based on the subsequent records, pCO2 in parts of the Oligocene was similar to latest Eocene values. These results show that a decrease in pCO2 preceded the large shift in marine oxygen isotope records that characterizes the Eocene–Oliogocene transition. This may be related to the "hysteresis effect" previously proposed – where a certain threshold of pCO2 change was crossed before the cumulative effects of this and other factors resulted in rapid temperature decline, ice build up on Antarctica and hence a change of climate mode.

Neural Machine Translation (NMT) models achieve their best performance when large sets of parallel data are used for training. Consequently, techniques for augmenting the training set have become popular recently. One of these methods is back-translation (Sennrich et al., 2016), which consists on generating synthetic sentences by translating a set of monolingual, target-language sentences using a Machine Translation (MT) model. Generally, NMT models are used for back-translation. In this work, we analyze the performance of models when the training data is extended with synthetic data using different MT approaches. In particular we investigate back-translated data generated not only by NMT but also by Statistical Machine Translation (SMT) models and combinations of both. The results reveal that the models achieve the best performances when the training set is augmented with back-translated data created by merging different MT approaches.

Multilingualism is a cultural cornerstone of Europe and firmly anchored in the European treaties including full language equality. However, language barriers impacting business, cross-lingual and cross-cultural communication are still omnipresent. Language Technologies (LTs) are a powerful means to break down these barriers. While the last decade has seen various initiatives that created a multitude of approaches and technologies tailored to Europe’s specific needs, there is still an immense level of fragmentation. At the same time, AI has become an increasingly important concept in the European Information and Communication Technology area. For a few years now, AI – including many opportunities, synergies but also misconceptions – has been overshadowing every other topic. We present an overview of the European LT landscape, describing funding programmes, activities, actions and challenges in the different countries with regard to LT, including the current state of play in industry and the LT market. We present a brief overview of the main LT-related activities on the EU level in the last ten years and develop strategic guidance with regard to four key dimensions. Peer reviewed

Natural language processing (NLP) can be done using either top-down (theory driven) and bottom-up (data driven) approaches, which we call mechanistic and phenomenological respectively. The approaches are frequently considered to stand in opposition to each other. Examining some recent approaches in deep learning we argue that deep neural networks incorporate both perspectives and, furthermore, that leveraging this aspect of deep learning may help in solving complex problems within language technology, such as modelling language and perception in the domain of spatial cognition. Comment: 18 pages, 1 figure, Appears in CLASP Papers in Computational Linguistics Vol. 1: Proceedings of the Conference on Logic and Machine Learning in Natural Language (LaML 2017)

This item is made available to you under the Creative Commons Attribution-Non commercial-No Derivatives 3.0 License. This paper aims to extend and update the survey of extreme wave events in Ireland that was previously carried out by O?Brien et al. (2013). The original catalogue highlighted the frequency of such events dating back as far as the turn of the last ice age through to 2012. Ireland?s marine territory extends far beyond its coastline and is one of the largest seabed territories in Europe. It is therefore not surprising that extreme waves have continued to occur regularly since 2012, particularly considering the severity of weather during the winters of 2013-14 and 2015-16. In addition, a large number of storm surges have been identified since the publication of the original catalogue. This paper updates the O?Brien et al. (2013) catalogue to include events up to the end of 2016. Storm surges are included as a new category and events are categorised into long waves (tsunamis and storm surges) and short waves (storm and rogue waves). New results prior to 2012 are also included and some of the events previously documented are reclassified. Important questions regarding public safety, services and the influence of climate change are also highlighted. Review Status: 27 Jul 2017 This discussion paper is a preprint. A revision of this manuscript was accepted for the journal Natural Hazards and Earth System Sciences (NHESS) and is expected to appear here in due course [https://www.nat-hazards-earth-syst-sci-discuss.net/nhess-2017-206/]

Curie depth offers a valuable constraint on the thermal structure of the lithosphere, based on its interpretation as the depth to 580 ∘C, but current methods underestimate the range of uncertainty. We formulate the estimation of Curie depth within a Bayesian framework to quantify its uncertainty across the British Isles. Uncertainty increases exponentially with Curie depth but this can be moderated by increasing the size of the spatial window taken from the magnetic anomaly. The choice of window size needed to resolve the magnetic thickness is often ambiguous but, based on our chosen spectral method, we determine that significant gains in precision can be obtained with window sizes 15–30 times larger than the deepest magnetic source. Our Curie depth map of the British Isles includes a combination of window sizes: smaller windows are used where the magnetic base is shallow to resolve small-scale features, and larger window sizes are used where the magnetic base is deep in order to improve precision. On average, the Curie depth increases from Laurentian crust (22.2±5.3 km) to Avalonian crust (31.2±9.2 km). The temperature distribution in the crust, and associated uncertainty, was simulated from the ensemble of Curie depth realizations assigned to a lower thermal boundary condition of a crustal model (sedimentary thickness, Moho depth, heat production, thermal conductivity), constructed from various geophysical and geochemical datasets. The uncertainty in the simulated heat flow field substantially increases from ±10 mW m−2 for shallow Curie depths at ∼15 km to ±80 mW m−2 for Curie depths >40 km. Surface heat flow observations are concordant with the simulated heat flow field except in regions that contain igneous bodies. Heat flow data within large batholiths in the British Isles exceed the simulated heat flow by ∼25 mW m−2 as a result of their high rates of heat production (4–6 µW m−3). Conversely, heat refraction around thermally resistive mafic volcanics and thick sedimentary layers induce a negative heat flow misfit of a similar magnitude. A northward thinning of the lithosphere is supported by shallower Curie depths on the northern side of the Iapetus Suture, which separates Laurentian and Avalonian terranes. Cenozoic volcanism in Northern Britain and Ireland has previously been attributed to a lateral branch of the proto-Icelandic mantle plume. Our results show that high surface heat flow (>90 mW m−2) and shallow Curie depth (∼15 km) occur within the same region, which supports the hypothesis that lithospheric thinning occurred due to the influence of a mantle plume. The fact that the uncertainty is only ±3–8 km in this region demonstrates that Curie depths are more reliable in hotter regions of the crust where the magnetic base is shallow.

We aim to shed light on the strengths and weaknesses of the newly introduced neural machine translation paradigm. To that end, we conduct a multifaceted evaluation in which we compare outputs produced by state-of-the-art neural machine translation and phrase-based machine translation systems for 9 language directions across a number of dimensions. Specifically, we measure the similarity of the outputs, their fluency and amount of reordering, the effect of sentence length and performance across different error categories. We find out that translations produced by neural machine translation systems are considerably different, more fluent and more accurate in terms of word order compared to those produced by phrase-based systems. Neural machine translation systems are also more accurate at producing inflected forms, but they perform poorly when translating very long sentences. Comment: Conference of the European Chapter of the Association for Computational Linguistics (EACL). April 2017, Val\`encia, Spain