• Digital Humanities and Cultural Heritage
• Publications close
• Other research products close
• US close
• Fraunhofer-ePrints close

ABSTRACTOoids are abundant carbonate grains throughout much of Earth's history, but their formation is not well understood. Here, an in‐depth study of microbial bioerosion features of Holocene ooids from the Schooner Cays ooid shoals (Great Bahama Bank, Eleuthera, Bahamas) and the Shalil al Ud ooid shoals in the Arabian/Persian Gulf (Abu Dhabi, United Arab Emirates) is presented. No obvious differences were found in ooid size distribution, cortex layer thickness, the composition of nuclei or euendolithic community when comparing ooids from both locations. Microendolithic borings are present in most studied ooid surfaces, but the intensity of (micro‐)bioerosion varies significantly. Applying an epoxy vacuum cast‐embedding technique allowed the identification of ichnotaxa and their inferred producers (various genera of diatoms, cyanobacteria, coccolithophores and unspecified bacteria). Euendolithic taxa have specific low‐light tolerances and light optima. This implies that information about the relative bathymetry (seafloor versus burial within an ooid shoal) and ecology for ooid cortex formation can be obtained via the presence or absence of their respective ichnotaxa. The history of a statistically significant number of ooid cortices can be translated into dune dynamics and the temporal variations thereof by allocating the inferred index producer to a defined burial or light penetration zone. In this context, ooid formation can be divided into four stages: (i) an agitation stage in the water column, characterized by the colonization of grains by photoautotrophs; (ii) a resting stage, characterized by temporary burial of the ooid, leading to immobilization and a shift towards heterotrophs; (iii) a sleeping stage, characterized by prolonged burial and colonization by organotrophs; and (iv) a reactivation stage, characterized by a resurfacing of the ooid and a subsequent shift towards photoautotrophs. The sleeping stage is presumably a stage of ooid degradation where bioerosion, mainly by heterotrophic fungi and bacteria is particularly active.

ABSTRACTMarine radiaxial fibrous calcites are common pore‐filling carbonate fabrics in Palaeozoic and Mesozoic carbonates, yet recent analogues are scarce. Although ancient marine radiaxial fibrous calcites are often considered a cement phase and used for palaeoenvironmental interpretations, their origin (primary precipitation versus diagenetic alteration of a pre‐existing fabric) and their specific formation processes remain ambiguous. This may be related to insufficient information from a few recent analogues and limitations of traditional methods in investigating ancient radiaxial fibrous calcites. This study documents a Holocene occurrence of marine pore‐filling radiaxial fibrous Mg calcites and focuses on their origin, formation processes and implications for ancient analogues. The radiaxial fibrous Mg calcites examined in this study share morphological and mineralogical similarities with numerous ancient marine radiaxial fibrous calcites with a Mg calcite precursor. Based on petrological and geochemical data, the radiaxial fibrous Mg calcites are interpreted as secondary porewater products formed during early burial alteration of micritic Mg calcite precursor cement. These findings provide implications for, at least in part, ancient marine radiaxial fibrous calcites in the context of the formation process and mechanism and are relevant for palaeoceanographic and palaeoenvironmental interpretations based on the presence of this carbonate phase.

While most approaches individually exploit unstructured data from the biomedical literature or structured data from biomedical knowledge graphs, their union can better exploit the advantages of such approaches, ultimately improving representations of biology. Using multimodal transformers for such purposes can improve performance on context dependent classification tasks, as demonstrated by our previous model, the Sophisticated Transformer Trained on Biomedical Text and Knowledge Graphs (STonKGs). In this work, we introduce ProtSTonKGs, a transformer aimed at learning all-encompassing representations of protein-protein interactions. ProtSTonKGs presents an extension to our previous work by adding textual protein descriptions and amino acid sequences (i.e., structural information) to the text- and knowledge graph-based input sequence used in STonKGs. We benchmark ProtSTonKGs against STonKGs, resulting in improved F1 scores by up to 0.066 (i.e., from 0.204 to 0.270) in several tasks such as predicting protein interactions in several contexts. Our work demonstrates how multimodal transformers can be used to integrate heterogeneous sources of information, paving the foundation for future approaches that use multiple modalities for biomedical applications.

MOTIVATION: The majority of biomedical knowledge is stored in structured databases or as unstructured text in scientific publications. This vast amount of information has led to numerous machine learning-based biological applications using either text through natural language processing (NLP) or structured data through knowledge graph embedding models. However, representations based on a single modality are inherently limited. RESULTS: To generate better representations of biological knowledge, we propose STonKGs, a Sophisticated Transformer trained on biomedical text and Knowledge Graphs (KGs). This multimodal Transformer uses combined input sequences of structured information from KGs and unstructured text data from biomedical literature to learn joint representations in a shared embedding space. First, we pre-trained STonKGs on a knowledge base assembled by the Integrated Network and Dynamical Reasoning Assembler consisting of millions of text-triple pairs extracted from biomedical literature by multiple NLP systems. Then, we benchmarked STonKGs against three baseline models trained on either one of the modalities (i.e. text or KG) across eight different classification tasks, each corresponding to a different biological application. Our results demonstrate that STonKGs outperforms both baselines, especially on the more challenging tasks with respect to the number of classes, improving upon the F1-score of the best baseline by up to 0.084 (i.e. from 0.881 to 0.965). Finally, our pre-trained model as well as the model architecture can be adapted to various other transfer learning applications. AVAILABILITY AND IMPLEMENTATION: We make the source code and the Python package of STonKGs available at GitHub (https://github.com/stonkgs/stonkgs) and PyPI (https://pypi.org/project/stonkgs/). The pre-trained STonKGs models and the task-specific classification models are respectively available at https://huggingface.co/stonkgs/stonkgs-150k and https://zenodo.org/communities/stonkgs. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online. ispartof: BIOINFORMATICS vol:38 issue:6 pages:1648-1656 ispartof: location:England status: published

In this paper, we propose an explainable approach for hate speech detection from the under-resourced Bengali language, which we called DeepHateExplainer. In our approach, Bengali texts are first comprehensively preprocessed, before classifying them into political, personal, geopolitical, and religious hates using a neural ensemble method of transformer-based neural architectures (i.e., monolingual Bangla BERT-base, multilingual BERT-cased/uncased, and XLM-RoBERTa). Subsequently, important (most and least) terms are identified using sensitivity analysis and layer-wise relevance propagation (LRP), before providing human-interpretable explanations11To foster reproducible research, we make available the data, source codes, models, and notebooks: https://github.com/rezacsedu/DeepHateExplainer. Finally, we compute comprehensiveness and sufficiency scores to measure the quality of explanations w.r.t faithfulness. Evaluations against machine learning (linear and tree-based models) and neural networks (i.e., CNN, Bi-LSTM, and Conv-LSTM with word embeddings) baselines yield F1-scores of 78%, 91%, 89%, and 84%, for political, personal, geopolitical, and religious hates, respectively, outperforming both ML and DNN baselines22Read an extended version of this paper: https://arxiv.org/abs/2012.14353.

The NEWA ALEX17 experiment was conducted with the objective of characterizing the wind conditions upstream of the Alaiz Test Site for the validation of flow models. From the intensive operational period, a case study has been selected for a Wakebench benchmark consisting of four consecutive days with relatively persistent winds from the North. The validation is centered around a 118-m mast at the Alaiz site and six additional masts located along the valley and at the lee side of a ridge delimiting a 8-km long area of interest. The benchmark is a follow-up of the GABLS3 diurnal cycle benchmark in flat terrain to test mesoscale-to-microscale transient and steady-state modeling methodologies in the assessment of stability-dependent bin-averaged wind conditions. Meso-micro methodologies reduce the wind speed mean bias from 32%, at 3-km mesoscale, to ±5%. Beyond mean bias mitigation, these initial results demonstrate the added value of meso-micro coupling at reproducing non conventional wind conditions at the test site like high-shear low-level jets in stable conditions and negative wind shear in unstable conditions. The benchmark also discusses the challenges of each meso-micro methodology going forward. This work was carried out under the umbrella of the International Energy Agency IEA-Wind TCP Task 31 Wakebench, with the support from: MARINE2 (EU grant agreement number 731084) UPWARDS (EU grant agreement number 763990), EoCoE-II (EU grant agreements number 824158 and 828947) and the Mexican Department of Energy, CONACYT-SENER Hidrocarburos grant agreement number B-S-69926 Peer Reviewed

In this paper, we propose CHOLAN, a modular approach to target end-to-end entity linking (EL) over knowledge bases. CHOLAN consists of a pipeline of two transformer-based models integrated sequentially to accomplish the EL task. The first transformer model identifies surface forms (entity mentions) in a given text. For each mention, a second transformer model is employed to classify the target entity among a predefined candidates list. The latter transformer is fed by an enriched context captured from the sentence (i.e. local context), and entity description gained from Wikipedia. Such external contexts have not been used in the state of the art EL approaches. Our empirical study was conducted on two well-known knowledge bases (i.e., Wikidata and Wikipedia). The empirical results suggest that CHOLAN outperforms state-of-the-art approaches on standard datasets such as CoNLL-AIDA, MSNBC, AQUAINT, ACE2004, and T-REx. Comment: accepted in EACL 2021 (full paper)

Pretrained Transformer models have emerged as state-of-the-art approaches that learn contextual information from text to improve the performance of several NLP tasks. These models, albeit powerful, still require specialized knowledge in specific scenarios. In this paper, we argue that context derived from a knowledge graph (in our case: Wikidata) provides enough signals to inform pretrained transformer models and improve their performance for named entity disambiguation (NED) on Wikidata KG. We further hypothesize that our proposed KG context can be standardized for Wikipedia, and we evaluate the impact of KG context on state-of-the-art NED model for the Wikipedia knowledge base. Our empirical results validate that the proposed KG context can be generalized (for Wikipedia), and providing KG context in transformer architectures considerably outperforms the existing baselines, including the vanilla transformer models. to appear in proceedings of CIKM 2020