Data from scientific studies are published in datasets, typically accompanied by data dictionaries and codebooks to support data understanding. To conduct rigorous analysis, data users need to leverage this documentation to correctly interpret the data. While this process can be burdensome for new data users, it is also prone to errors even for seasoned users. A computational formal model of the knowledge that was used to create the study can facilitate better understanding and thus improved usage of the study data. Knowledge graphs can be used efectively to capture this study knowledge. The SciKG tutorial aimed to introduce participants to the basics of knowledge graph construction using data, data dictionaries, and codebooks from scientific studies. It used the Center for Disease Control and Prevention's (CDC) National Health and Nutrition Examination Surveys (NHANES) data as a testbed and introduce standardized terminology, novel and established techniques, and resources such as scientific/biomedical ontologies, semantic data dictionaries, and knowledge graph frameworks in both lecture and practical sessions. Website: https://tetherless-world.github.io/scikg-eswc-2023/
The construction of knowledge graphs (KGs) for the biomedical domain (and, generally, the
scientific domain) is a prominent field, with much attention from the community. Several
venues have highlighted this, including the recently organized Personal Health Knowledge
Graph workshop1. Scientific KGs have been deployed to support increasing automation in
biomedical research, including for reproducible research [
SciKG was a full-day tutorial that introduced participants to the basics of knowledge graph
construction with input from datasets from scientific studies and surveys, as well as the
associated data dictionaries, codebooks, and documentation. To support this, the tutorial began
with an overview of state-of-the-art scientific and biomedical ontologies that are commonly
reused. Next, the participants were introduced to Semantic Data Dictionaries (SDDs) [
The SciKG tutorial was based on simple instructive (but insightful) examples. At the end of the tutorial, participants were be able to: • Identify and reuse relevant scientific and biomedical ontologies • Develop minimally working semantic data dictionaries that capture domain modeling from scientific data and documentation • Use knowledge graph frameworks to bootstrap and manage scientific knowledge graphs • Interact with the graph to retrieve data based on analysis-driven questions This tutorial covered methods and tools that are established and being used in production environments at several institutions, including National Institute of Health-funded projects at the Icahn School of Medicine at Mount Sinai2, McGill University’s Peter Guo-hua Fu School of Architecture3, Rensselaer Polytechnic Institute’s Tetherless World Constellation4, and Escola de Ciência de Informação at Universidade Federal de Minas Gerais5. The SciKG tutorial was inspired by previous tutorials on knowledge graph construction, including the Knowledge Graph Construction Tutorial6 held at ESWC 2022, and the Tools for Creating and Exploiting Large Knowledge Graphs (KGTK)7 held at ISWC 2021. These tutorials were focused on general tools for KG building, management, and exploration. SciKG, in its turn, focused on the construction of KGs from scientific data, with rigor in scientific knowledge maintenance and representation, covering not only techniques, but standardized scientific terminology and best practices. According to the conference organizers, SciKG had an average attendance of 25 people.
SciKG was primarily targeted at Semantic Web researchers working (or willing to work) with biomedical data. However, the acquired knowledge applies to studies beyond the biomedical domain, as the aspects of the scientific methods for data acquisition are common. The target audience included students and researchers interested in bioinformatics settings, as well as anyone interested in conducting research using scientific data, often from multiple sources.
The SciKG tutorial was divided into four sections. It started with an overview of how scientific study data is usually acquired, organized and published, and the current challenges (and opportunities for semantic web) involving the use of this data. Next, we introduced methods for scientific data annotation and terminology reuse. Following, we gave an overview of the state-of-the-art scientific and biomedical ontologies and provided real-world examples of their successful adoption. Finally, the tutorial introduced knowledge graph frameworks and demonstrate how they can be used to bootstrap and manage scientific KGs.
3https://www.mcgill.ca/architecture/ 4https://tw.rpi.edu/project/human-health-exposure-analysis-repository-hhear 5http://eci.ufmg.br 6https://kg-construct.github.io/eswc-dkg-tutorial-2022/ 7https://usc-isi-i2.github.io/kgtk-tutorial-iswc-2021/ This section presented a common scientific methodology used in scientific studies and demonstrate how study data is usually acquired, organized, and published. Using the NHANES survey as an example, this section demonstrated how study documentation is usually used to describe the contents of data files using data dictionaries and codebooks, and how it is problematic in capturing all the semantics associated with the variable. Data dictionaries document the study variables contained in data files, providing a natural language description of the variable. For instance, the RIDAGEYR variable in NHANES is defined as Age in years of the participant at the time of screening. Individuals 80 and over are topcoded at 80 years of age, and we can infer that it is measuring the age of the survey participant (not of other persons included in the data), using years as the unit of measurement (not months which is used for infants), recorded during screening time (instead of examination time).
This section introduced participants to some well-used state-of-the-art scientific and biomedical
ontologies, including the Semanticscience Integrated Ontology (SIO) [
This topic demonstrated techniques for capturing study knowledge from data, and
documentation. We introduced the Semantic Data Dictionary method for aligning and integrating data and
provide a suficient set of examples. A Semantic Data Dictionary (SDD) is a model for
representing metadata from ontologies and structured vocabularies through a set of specifications that
allows the assignment of a semantic representation of the data [
This topic introduced knowledge graph frameworks, and presented two that have built-in
capabilities of working with SDDs, including the Human-Aware Data Acquisition Framework
(HADatAc) [
Tutorial website NHANES SDDs HADatAc Whyis