Semantic Framework for Practicing Data Science in Public Health Organizations during the Covid-19 Pandemics

Abstract

This paper proposes a semantic framework based on software architectures for accommodating data science practices to the needs of Public Health Organizations (PHO), during the covid-19 pandemics. The goal is to create an environment suitable for deploying data science on an ad-hoc basis, upon the selection of data generated by governments, nongovernment organizations, public databases and social media, but guided by PHO own needs and expertise. It is important to run predictions, through learning technologies, which may depend on circumstances and situations relevant for PHO in the particular moment and thus enable better decision making in the time of the pandemic. The proposed software architecture relies on its deployment within integrated development environments and plug-ins/APIs towards software tools, and libraries for (a) data gathering and preprocessing, (b) predictions with learning technologies (c) reasoning with semantic technologies and (d) including human intervention to aid in understanding the situation in which PHO questions may be answered. The illustration of the proposal is uses the sentiment analysis of twitter data relevant to covid-19 and classification of tweets with machine learning.

EDGIE: A simulation test-bed for investigating the impacts of building and vehicle electrification on distribution grids

“…This may deliver the promise of addressing production costs and the burden of complicated and expensive procedures of bringing new drugs to the market. It is premature to say that ML algorithms are sine qua non for a new era of drug discoveries and repurposing with computational polypharmacology (Juric and Ronchieri, 2022), (Juric et al, 2020), (Almami et al, 2016a).…”

“…This means that we cannot place all these repositories in one "basket" and assume that semantics stored in them would be computable (Juric et al, 2021). Creating centralized repositories for excessive amounts of data, with complex semantics, is a no-go area for efficient computational models today (Juric, 2022). The time of complex and centralized software solutions are long gone.…”

“…We accept that predictions are derived from statistical models, available within learning technologies, known as ML algorithms, which shape current AI. However, reasoning and predicting are two different types of computations, and they cannot be mixed and matched without having a specific software architectural solution (Juric and Ronchieri, 2022), (Juric and Kim, 2017). Therefore, advances in predictions of drug combination based on clinical sideeffects (Huang, et al, 2014) and ML in drug discoveries (Talevi et al, 2020), predictions of synergistic drug combinations (Gayvert et al, 2017), and selective combinations of druggable targets (Tang et al, 2013) are examples of "predictions".…”

Does AI Disclosure in Discriminatory Pricing Backfire? The Moderating Role of Price Sensitivity and Explanation of Price Differences

“…The interest of merging the two for creating an ECP is triggered by successes in using the SWT for creating computational models, powered by SWRL enabled OWL ontologies (Juric, 2016) (Juric and Kim, 2017), (Shojanoori and Juric, 2014).. At this moment there is nothing else on offer in either computer science or communication theories or predictive technologies or engineering which could take us forward to a fully functional ECP. This is particularly true in cases of instances of the IoT.…”