Computational Thinking
We choose to deploy constructs from Computational Thinking (CT) in the context of linked data projects specifically with a mind toward teaching students how to think about the scaling problems when attempting to maximally organize sets of resources. We have two active projects in the Linked Data Research Group, one of which is well developed and will serve as our primary example to students and the second project is at a more nascent stage. They are:
- Organizing access to multimedia (still images and video clips) that document game action in college football games using data sources internal to a football game (e.g., statistical play-by-play datasets) and those data sources that are external (e.g., transcripts of television broadcasts).
- Organizing access to textual materials of the HathiTrust page scans by examining back-of-book indexes and ToCs as access mechanisms for internal book content. Also capturing features of interest to descriptive bibliographers.
The CT constructs below were adapted from Computational Practices Adapted to Archival Science work at the University of Maryland iSchool and are aimed at delineating CT processes needed to understand the Link Data Research Group's work on project 1 above.
We have chosen to distill our research and teaching efforts for using linked data to organize resource collections into the following two themes:
- Data-driven semantic indexing over RDF graphs
- Our data is metadata
Both themes can be operationalized using an extract metadata/transform semantically/load (ETL) workflow framework to further characterize the CT constructs selected from the University of Maryland iSchool model for archival science.
High-level Description of Project
(For additional detailed information and background on this project, see Chronology: Data-driven Sports Image Indexing Research and sections of my 2018-19 Academic Year Research Report)
How can we index individual images that document football game action at scale? There are so many teams, so many photographers, so many photographs, all problems compounded by two material states of images (born digital and film photographs). Content-based computational methods are available to analyze images; however, these approaches alone are subject to semantic gap problems.
