To investigate the cause and progression of a phenomenon, such as chronic disease, it is essential to collect a wide variety of data that together explains the complex interplay of di erent factors, e.g., genetic, lifestyle, environmental and social. Sharing information between studies is therefore of paramount importance. However, data that needs to be analyzed must be appropriately integrated, conceptually aligned, and harmonized. This implies that data collection must be done either in a su ciently similar or a su ciently transparent way in order to support meaningful synthesis from di erent studies. We will demonstrate4,5 how the Human-Aware Data Acquisition (HADatAc) framework integrates and harmonizes data from multiple scienti c studies and thus how to use it in interdisciplinary science investigations.
The Human-Aware Data Acquisition (HADatAc) Framework is a schema-free,
evolutionary, scalable and provenance-aware infrastructure for managing data
and metadata content from multiple scienti c studies. Three key goals of
HADatAc are: (
Without any expansion, the framework and the collection of ontologies listed above are domain agnostic and ready for usage in scienti c domains. Existing HADatAc deployments are built using these core ontologies, along with many other specialized ontologies, for the domain of interest. One speci c domain ontology is often used to import specialized ontologies into a single document that HADatAc uses as a default namespace for a given domain of interest (e.g. the integrated exposure and health ontology CHEAR [3] that we use in our HADatAc backend for the NIEHS Child Health Exposure Analysis Resource implementation). 2
Schema-free claim. HADatAc uses semantic technologies, ontologies, graph databases and non-relational databases to manage metadata and data from relevant scienti c studies, e.g. CDC's National Health and Nutrition Survey (NHANES)7 in the demonstration system. HADatAc is schema-free since the content from these study les is stored without a prede ned and xed structure. This allows HADatAc to include objects, e.g., subjects, samples, locations, into its repositories as they are presented, including attributes from objects regarded as relevant for the studies.
Evolutionary claim. Ontologies and the underlying graph, including the loaded data, are managed by one Apache SOLR repository and by one Blazegraph RDF graph database. The SOLR repository manages data values and a collection of URIs for each data value. URIs, in the collection of URIs of each stored data value, are links between the data value and semantic annotations in the Blazegraph. The Blazegraph repository is used to manage the HADatAc knowledge graph. HADatAc's knowledge graph evolves by either importing existing ontologies or by de ning concepts and relations that are not available or not appropriate for reuse from existing ontologies.
Scalability claim. Scienti c data management platforms must handle increasing data volumes. HADatAc's backend SOLR repository provides the required scalability for very large data repositories. As the metadata volume is signi cantly smaller than the data volume, it is easily managed by the Blazegraph triple-store. SOLR is also used to compute aggregate faceted values into the scienti c study indicators, that would be costly to do with Blazegraph.
Provenance-Aware claim. HADatAc has been developed to work with a broad range of data sources, and the infrastructure captures and preserves provenance on how data values where acquired by a broad notion of instruments. HADatAc classi es data sources according to the instruments (and detectors) used to acquire the data. HADatAc understands if study data is the result of any of the data acquisition strategies: (a) empirical measurement, which is done using physical instruments like sensors; (b) data and knowledge elicitation from humans, which is done using questionnaires as instruments; (c) computer data generation (simulation), which is done using simulation models as instruments. 7https://www.cdc.gov/nchs/nhanes/index.htm
Metadata file
File Management
Data files
Study Management
Instrument
Management SOLR (data)
Blazegraph (metadata)
HAScO POJO
Library Metadata Data Ingestion Ingestion
Content Ingestion Core Component
Search Download Alignment
API
HADatAc is a framework and also a web application. Most of its web user interface is implemented as part of the six satellite subsystems. The API Subsystem is a special subsystem composed of a collection of RESTful services with programmatic access to HADatAc's content. The Core Component has the elements required to support the satellite subsystems including: the SOLR and Blazegraph content repositories, a Java API encoding the concepts of the Human-Aware Science Ontology (HAScO) as POJO Classes, and the subsystems responsible for extracting, annotating, and storing study content from data and metadata les into SOLR and Blazegraph. The HAScO POJO classes are used to build and maintain HADatAc's knowledge graph.
Content is added into HADatAc either through the parsing of uploaded les through the File Management Subsystem or on-line through user interaction with the Study Management Subsystem and the Instrument Management Subsystem. Content is directly presented to users through the Search Subsystem and downloaded through the Object Alignment Subsystem and API Subsystem. 4
Our demonstration includes the following six steps: (