Collected in this dataset are the slideset and abstract for a presentation on Toward a Reproducible Research Data Repository by the depositar team at International Symposium on Data Science 2023 (DSWS 2023), hosted by the Science Council of Japan in Tokyo on December 13-15, 2023. The conference was organized by the Joint Support-Center for Data Science Research (DS), Research Organization of Information and Systems (ROIS) and the Committee of International Collaborations on Data Science, Science Council of Japan. The conference programme is also included as a reference.
Title
Toward a Reproducible Research Data Repository
Author(s)
Cheng-Jen Lee, Chia-Hsun Ally Wang, Ming-Syuan Ho, and Tyng-Ruey Chuang
Affiliation of presenter
Institute of Information Science, Academia Sinica, Taiwan
Summary of Abstract
The depositar (https://data.depositar.io/) is a research data repository at Academia Sinica (Taiwan) open to researhers worldwide for the deposit, discovery, and reuse of datasets. The depositar software itself is open source and builds on top of CKAN. CKAN, an open source project initiated by the Open Knowledge Foundation and sustained by an active user community, is a leading data management system for building data hubs and portals. In addition to CKAN's out-of-the-box features such as JSON data API and in-browser preview of uploaded data, we have added several features to the depositar, including sourcing from Wikidata as dataset keywords, a citation snippet for datasets, in-browser Shapefile preview, and a persistent identifier system based on ARK (Archival Resource Keys). At the same time, the depositar team faces an increasing demand for interactive computing (e.g. Jupyter Notebook) which facilitates not just data analysis, but also for the replication and demonstration of scientific studies. Recently, we have provided a JupyterHub service (a multi-tenancy JupyterLab) to some of the depositar's users. However, it still requires users to first download the data files (or copy the URLs of the files) from the depositar, then upload the data files (or paste the URLs) to the Jupyter notebooks for analysis. Furthermore, a JupyterHub deployed on a single server is limited by its processing power which may lower the service level to the users. To address the above issues, we are integrating the BinderHub into the depositar. BinderHub (https://binderhub.readthedocs.io/) is a kubernetes-based service that allows users to create interactive computing environments from code repositories. Once the integration is completed, users will be able to launch Jupyter Notebooks to perform data analysis and vsualization without leaving the depositar by clicking the BinderHub buttons on the datasets. In this presentation, we will first make a brief introduction to the depositar and BinderHub along with their relationship, then we will share our experiences in incorporating interactive computation in a data repository. We shall also evaluate the possibility of integrating the depositar with other automation frameworks (e.g. the Snakemake workflow management system) in order to enable users to reproduce data analysis.
Keywords
BinderHub, CKAN, Data Repositories, Interactive Computing, Reproducible Research
