ABSTRACT The wide applications of Generative adversarial networks benefit from the successful training methods, guaranteeing that an object function converges to the local minimum. Nevertheless, designing an efficient and competitive training method is still a challenging task due to the cyclic behaviors of some gradient-based ways and the expensive computational cost of acquiring the Hessian matrix. To address this problem, we proposed the Adaptive Composite Gradients(ACG) method, linearly convergent in bilinear games under suitable settings. Theory analysis and toy-function experiments both suggest that our approach alleviates the cyclic behaviors and converges faster than recently proposed SOTA algorithms. The convergence speed of the ACG is improved by 33% than other methods. Our ACG method is a novel Semi-Gradient-Free algorithm that can reduce the computational cost of gradient and Hessian by utilizing the predictive information in future iterations. The mixture of Gaussians experiments and real-world digital image generative experiments show that our ACG method outperforms several existing technologies, illustrating the superiority and efficacy of our method.

Abstract Timestamps play a key role in process mining because it determines the chronology of which events occurred and subsequently how they are ordered in process modelling. The timestamp in process mining gives an insight on process performance, conformance, and modelling. This therefore means problems with the timestamp will result in misrepresentations of the mined process. A few articles have been published on the quantification of data quality problems but just one of the articles at the time of this paper is based on the quantification of timestamp quality problems. This article evaluates the quality of timestamps in event log across two axes using eleven quality dimensions and four levels of potential data quality problems. The eleven data quality dimensions were obtained by doing a thorough literature review of more than fifty process mining articles which focus on quality dimensions. This evaluation resulted in twelve data quality quantification metrics and the metrics were applied to the MIMIC-III dataset as an illustration. The outcome of the timestamp quality quantification using the proposed typology enabled the user to appreciate the quality of the event log and thus makes it possible to evaluate the risk of carrying out specific data cleaning measures to improve the process mining outcome.

ABSTRACT Functional Status Information (FSI) describes physical and mental wellness at the whole-person level. It includes information on activity performance, social role participation, and environmental and personal factors that affect the well-being and quality of life. Collecting and analyzing this information is critical to address the needs for caring for an aging global population, and to provide effective care for individuals with chronic conditions, multi-morbidity, and disability. Personal knowledge graphs (PKGs) represent a suitable way for meaning in a complete and structured way all information related to people's FSI and reasoning over them to build tailored coaching solutions supporting them in daily life for conducting a healthy living. In this paper, we present the development process related to the creation of a PKG by starting from the HeLiS ontology in order to enable the design of an AI-enabled system with the aim of increasing, within people, the self-awareness of their own functional status. In particular, we focus on the three modules extending the HeLiS ontology aiming to represent (i) enablers and (ii) barriers playing potential roles in improving (or deteriorating) own functional status and (iii) arguments driving the FSI collection process. Finally, we show how these modules have been instantiated into real-world scenarios.

ABSTRACT To alleviate the problem of under-utilization features of sentence-level relation extraction, which leads to insufficient performance of the pre-trained language model and underutilization of the feature vector, a sentence-level relation extraction method based on adding prompt information and feature reuse is proposed. At first, in addition to the pair of nominals and sentence information, a piece of prompt information is added, and the overall feature information consists of sentence information, entity pair information, and prompt information, and then the features are encoded by the pre-trained language model ROBERTA. Moreover, in the pre-trained language model, BIGRU is also introduced in the composition of the neural network to extract information, and the feature information is passed through the neural network to form several sets of feature vectors. After that, these feature vectors are reused in different combinations to form multiple outputs, and the outputs are aggregated using ensemble-learning soft voting to perform relation classification. In addition to this, the sum of cross-entropy, KL divergence, and negative log-likelihood loss is used as the final loss function in this paper. In the comparison experiments, the model based on adding prompt information and feature reuse achieved higher results of the SemEval-2010 task 8 relational dataset.

ABSTRACT In recent years, feature engineering-based machine learning models have made significant progress in auto insurance fraud detection. However, most models or systems focused only on structural data and did not utilize multi-modal data to improve fraud detection efficiency. To solve this problem, we adapt both natural language processing and computer vision techniques to our knowledge-based algorithm and construct an Auto Insurance Multi-modal Learning (AIML) framework. We then apply AIML to detect fraud behavior in auto insurance cases with data from real scenarios and conduct experiments to examine the improvement in model performance with multi-modal data compared to baseline model with structural data only. A self-designed Semi-Auto Feature Engineer (SAFE) algorithm to process auto insurance data and a visual data processing framework are embedded within AIML. Results show that AIML substantially improves the model performance in detecting fraud behavior compared to models that only use structural data.

Abstract This study explored the regulatory framework in Kenya that may facilitate the implementation of the FAIR Guidelines in health research, as well as the possibility of adopting the FAIR Guidelines at the national level. Fourteen key documents pivotal to the emerging digital health sector in Kenya were identified and analysed using a comprehensive coding and labelling approach based on a binary system for whether or not they mention the FAIR Guidelines or terms and vocabulary related to the FAIR Guidelines. The analysis revealed gaps in data stewardship that could be filled by the implementation of the FAIR Guidelines and, although the documents analysed do not explicitly mention the FAIR Guidelines, FAIR Equivalent terminology and practices are mentioned in varying detail. However, our analysis shows that there are still no provisions for the introduction and implementation of the FAIR Guidelines in health research in Kenya. Therefore, we recommend that the leadership be provided with a comprehensive introduction to the FAIR Guidelines, success stories about the FAIRification of data and research infrastructure in other parts of the world, and a demonstration of the steps needed for the FAIRification of health data in Kenya.

Abstract This article describes the FAIRification process (which involves making data Findable, Accessible, Interoperable and Reusable—or FAIR—for both machines and humans) for data related to the impact of COVID-19 on migrants, refugees and asylum seekers in Tunisia, Libya and Niger, according to the scheme adopted by GO FAIR. This process was divided into three phases: pre-FAIRification, FAIRification and post-FAIRification. Each phase consisted of seven steps. In the first phase, 118 in-depth interviews and 565 press articles and research reports were collected by students and researchers at the University of Sousse in Tunisia and researchers in Niger. These interviews, articles and reports constitute the dataset for this research. In the second phase, the data were sorted and converted into a machine actionable format and published on a FAIR Data Point hosted at the University of Sousse. In the third phase, an assessment of the implementation of the FAIR Guidelines was undertaken. Certain barriers and challenges were faced in this process and solutions were found. For FAIR data curation, certain changes need to be made to the technical process. People need to be convinced to make these changes and that the implementation of FAIR will generate a long-term return on investment. Although the implementation of FAIR Guidelines is not straightforward, making our resources FAIR is essential to achieving better science together.

Abstract Temporal information is pervasive and crucial in medical records and other clinical text, as it formulates the development process of medical conditions and is vital for clinical decision making. However, providing a holistic knowledge representation and reasoning framework for various time expressions in the clinical text is challenging. In order to capture complex temporal semantics in clinical text, we propose a novel Clinical Time Ontology (CTO) as an extension from OWL framework. More specifically, we identified eight time-related problems in clinical text and created 11 core temporal classes to conceptualize the fuzzy time, cyclic time, irregular time, negations and other complex aspects of clinical time. Then, we extended Allen's and TEO's temporal relations and defined the relation concept description between complex and simple time. Simultaneously, we provided a formulaic and graphical presentation of complex time and complex time relationships. We carried out empirical study on the expressiveness and usability of CTO using real-world healthcare datasets. Finally, experiment results demonstrate that CTO could faithfully represent and reason over 93% of the temporal expressions, and it can cover a wider range of time-related classes in clinical domain.

Abstract Data harmonization and documentation of the data processing are essential prerequisites for enabling Canonical Analysis Workflows. The recently revised Terabyte-scale air quality database system, which the Tropospheric Ozone Assessment Report (TOAR) created, contains one of the world's largest collections of near-surface air quality measurements and considers FAIR data principles as an integral part. A special feature of our data service is the on-demand processing and product generation of several air quality metrics directly from the underlying database. In this paper, we show that the necessary data harmonization for establishing such online analysis services goes much deeper than the obvious issues of common data formats, variable names, and measurement units, and we explore how the generation of FAIR Digital Objects (FDO) in combination with automatically generated documentation may support Canonical Analysis Workflows for air quality and related data.

Abstract In this paper we present the Reproducible Research Publication Workflow (RRPW) as an example of how generic canonical workflows can be applied to a specific context. The RRPW includes essential steps between submission and final publication of the manuscript and the research artefacts (i.e., data, code, etc.) that underlie the scholarly claims in the manuscript. A key aspect of the RRPW is the inclusion of artefact review and metadata creation as part of the publication workflow. The paper discusses a formalized technical structure around a set of canonical steps which helps codify and standardize the process for researchers, curators, and publishers. The proposed application of canonical workflows can help achieve the goals of improved transparency and reproducibility, increase FAIR compliance of all research artefacts at all steps, and facilitate better exchange of annotated and machine-readable metadata.