International audience

In geoscience, it is necessary to study the lithography of the Earth's subsurface, which consists of different stratified layers called geological formations.This study performs well correlation task to model and characterize reservoirs.This operation links the beginning of specific geological formations called tops using measurements from drilled wells.Although data are abundant, the traditional algorithms used for well correlation are semi-automated, requiring significant time and high computational cost.This paper introduces GeoTS, a Python library to apply cutting-edge time series classification models to perform well correlation in a completely automated setting.As input, take the drilling trajectory depth and gamma-ray well logs, which measure the natural radioactivity across the well depth trajectory.The top depths of the formations are predicted as an output.The gamma-ray signatures are extracted around the top depths assigned by geologists.Preprocessing is performed to clean and cluster these signatures using the Dynamic Time Wrapping (DTW) distance and HDBSCAN.Implementation of existing deep learning architectures (FCN, InceptionTime, XceptionTime, XCM, LSTM-FCN) and new architecture (LSTM-2dCNN, LSTM-XCM) are performed.Our experiments demonstrate faster computation with an increase in accuracy.GradCAM has also been implemented for model explainability.Experiments were performed using Colorado oil fields and deployed on Wyoming oil fields.The deployment has provided us with critical insights regarding the improvements needed.

Manually integrating data of diverse formats and languages is vital to many artificial intelligence applications. However, the task itself remains challenging and time-consuming. This paper highlights the potential of Large Language Models (LLMs) to streamline data extraction and resolution processes. Our approach aims to address the ongoing challenge of integrating heterogeneous data sources, encouraging advancements in the field of data engineering. Applied on the specific use case of learning disorders in higher education, our research demonstrates LLMs’ capability to effectively extract data from unstructured sources. It is then further highlighted that LLMs can enhance data integration by providing the ability to resolve entities originating from multiple data sources. Crucially, the paper underscores the necessity of preliminary data modeling decisions to ensure the success of such technological applications. By merging human expertise with LLM-driven automation, this study advocates for the further exploration of semi-autonomous data engineering pipelines.

Cities serve as vital hubs of economic activity and knowledge generation and dissemination. As such, cities bear a significant responsibility to uphold environmental protection measures while promoting the welfare and living comfort of their residents. There are diverse views on the development of smart cities, from integrating Information and Communication Technologies into urban environments for better operational decisions to supporting sustainability, wealth, and comfort of people. However, for all these cases, data are the key ingredient and enabler for the vision and realization of smart cities. This article explores the challenges associated with smart city data. We start with gaining an understanding of the concept of a smart city, how to measure that the city is a smart one, and what architectures and platforms exist to develop one. Afterwards, we research the challenges associated with the data of the cities, including availability, heterogeneity, management, analysis, privacy, and security. Finally, we discuss ethical issues. This article aims to serve as a “one-stop shop” covering data-related issues of smart cities with references for diving deeper into particular topics of interest.

This paper addresses the need to better support dyslexic students in higher education using data-driven methods. Our approach lies in modeling relationships between dyslexic students, problems inherent to their condition, and potential solutions. The proposed graph database integrates multiple data sources, in different formats and languages, and captures complex relationships between entities that are not identifiable when considering each data source independently. This paper’s main contribution is a hybrid recommender system that first filters potential solutions through the navigation of the modeled graph, then utilizes a Neural Network to solve an ordinal classification problem: effectively ranking the filtered recommendations based on their predicted usefulness. Several documented approaches to solving the ranking algorithm’s prediction task were implemented and compared. The models that achieved the highest ranking accuracy, approximately 74%, were 3-Layer Neural Networks trained with Ordinal Log-Loss and self-guided EMD2loss. In summary, our work not only facilitates the identification of patterns essential for crafting personalized recommendations to address the most severe difficulties of dyslexic students, but also establishes a structured foundation for the scalable integration of additional data sources. These results strongly support future research and application development related to dyslexia in higher education.

The global health crisis represents an unprecedented opportunity for the development of artificial intelligence (AI) solutions. This article aims to tackle part of the biases in artificial intelligence by implementing a human-centric AI to help decision-makers in organizations. It relies on the results of two design science research (DSR) projects: SCHOPPER and VRAILEXIA. These two design projects operationalize the human-centric AI approach with two complementary stages: 1) the first installs a human-in-loop informed design process, and 2) the second implements a usage architecture that aggregates AI and humans. The proposed framework offers many advantages such as permitting to integrate of human knowledge into the design and training of the AI, providing humans with an understandable explanation of their predictions, and driving the advent of augmented intelligence that can turn algorithms into a powerful counterweight to human decision-making errors and humans as a counterweight to AI biases.

Twitter is a social network that provides a powerful source of data. The analysis of those data offers many challenges among those stands out the opportunity to find reputation of a product, a person or any other entity of interest. Several approaches for sentiment analysis have been proposed in the literature to assess the general opinion expressed in tweets on an entity. Nevertheless, these methods aggregate sentiment scores retrieved from tweets, which is a static view to evaluate the overall reputation of an entity. The reputation of an entity is not static; entities collaborate with each other, and they get involved in different events over time. A simple aggregation of sentiment scores is then not sufficient to represent this dynamism. In this paper, we present a new approach to determine the reputation of an entity on the basis of the set of events in which it is involved. To achieve this, we propose a new sampling method driven by a tweet weighting measure to give a better quality and summary of the target entity. We introduce the concept of Frequent Named Entities to determine the events involving the target entity. Our evaluation achieved for different entities shows that 90% of the reputation of an entity originates from the events it is involved in and the breakdown into events allows interpreting the reputation in a transparent and self-explanatory way.

Data management goes through interesting times1, as the number of currently available data management systems (DMSs in short) is probably higher than ever before. This leads to unique opportunities for data-intensive applications, as some systems provide excellent performance on certain data processing operations. Yet, it also raises great challenges, as a system efficient on some tasks may perform poorly or not support other tasks, making it impossible to use a single DMS for a given application. It is thus desirable to use different DMSs side by side in order to take advantage of their best performance, as advocated under terms such as hybrid or poly-stores. We present ESTOCADA, a novel system capable of exploiting side-by-side a practically unbound variety of DMSs, all the while guaranteeing the soundness and completeness of the store, and striving to extract the best performance out of the various DMSs. Our system leverages recent advances in the area of query rewriting under constraints, which we use to capture the various data models and describe the fragments each DMS stores.

Data processing is the core of any statistical information system. Statisticians are interested in specifying transformations and manipulations of data at a high level, in terms of entities of statistical models such as time series. We illustrate here an experience at the Bank of Italy where (i) a language, EXL, has been defined for the declarative specification of statistical programs, (ii) an approach for the translation of EXL code into executables in various target systems has been developed, and (iii) a concrete implementation, EXLEngine, has been carried out. The approach leverages on schema mappings as an intermediate specification step, in order to facilitate the translation from EXL towards several target systems.

We present AMADA, a platform for storing Web data (in particular, XML documents and RDF graphs) based on the Amazon Web Services (AWS) cloud infrastructure. AMADA operates in a Software as a Service (SaaS) approach, allowing users to upload, index, store, and query large volumes of Web data. The demonstration shows (i) the step-by-step procedure for building and exploiting the warehouse (storing, indexing, querying) and (ii) the monitoring tools enabling one to control the expenses (monetary costs) charged by AWS for the operations involved while running AMADA.

The recent growth of non-relational databases (often termed as NoSQL) is an interesting phenomenon that has generated both interest and criticism. One of the major drawbacks that is often referred to is the heterogeneity of the languages and interfaces they offer to developers and users.

A runtime approach to model-generic translation of schema and data is proposed.