New Perspectives on Your Data Using Heterogeneous Information Networks (NFDI4Objects Community Meeting 2024)

New Perspectives on Your Data Using Heterogeneous Information Networks HIN Creation: Accurately Representing Information HINs aim to represent heterogeneous data in an information-rich, semantically correct way, thus enabling successful analysis to obtain new information about the modeled data. Creating an effective HIN requires close collaboration between domain experts and computer scientists to accurately transform tabular data into a network structure. Nodes represent unique combinations of feature values, with edges representing their semantic relationships. Standardized ontologies, along with authority files and controlled vocabularies, are tools for consistent modeling of diverse datasets as HINs. Through these standards, successful integration of datasets can be ensured, leading to an enriched knowledge base. For details on HINs, see Shi et al. (2017). Visual HIN Creation Process: From raw data to an integrated network. Bibliography: •Schmid, C. et al. (2024). https://doi.org/10.1101/2024.04.12.589180 •Shi, C. et al. (2017). https://doi.org/10.1109/TKDE.2016.2598561 •Rossi, R. A. et al. (2020). https://doi.org/10.1145/3418773 •Sun, Y. et al. (2013). https://doi.org/10.14778/3402707.3402736 Imprint: NFDI4Objects Community Meeting 2024. 25.-27. September 2024 Christian-Albrechts-Universität zu Kiel Department of Computer Science Working Group: Archaeoinformatic –Data Science Christian-Albrechts-Platz 4 24118 Kiel Contact Details: Interested? Get in touch. [email protected] LinkedIn Mattis thor Straten Matthias Renz Department of Computer Science, Kiel University Use-Case: Transforming the Poseidon Community Archive The Poseidon Community Archive stores publication-wise genotype data. Each publication contains a file consisting of archaeological context and metadata for genotyped individuals. We use a directed, typed graph structure—a Heterogeneous Information Network (HIN)—to accurately represent the information-rich, integrated data. This network representation offers a novel data view, enabling rich analysis leading to the potential discovery of new insights. For details on Poseidon, see Schmid et al. (2024). Transformation and integration of tabular data files into an integrated Heterogeneous Information Network. Which material is more relevant to Bone material? Relevance Search identifies the most relevant objects given a query object. It can be used as a stand-alone analysis or for tasks like link prediction and recommendation. Clustering groups objects into meaningful subsets (clusters), ensuring high intra-cluster similarity and low inter-cluster similarity. Which objects form groups? Network Motifs are small subgraphs that occur frequently within the network. They provide critical insights into the network’s structural design and reveal hidden characteristics. For details, see Rossi et al. (2020). Which subgraphs occur frequently? Exemplary meta paths on the Poseidon network connecting two materials related to samples with similar site information. HIN Analysis: Uncovering Hidden Insights HINs offer a new link-based perspective on the modeled data. Analysis may reveal hidden insights by examining the network structure. Meta paths—sequences of relationship types—capture complex higher-order relationships between object types in the network. Each meta path represents a specific semantic relationship and can be used for network analysis. For details, see Sun et al. (2013). Material images generated using ChatGPT