International Data Week - Random Notes

[joehand]

I'm at international data week this week. Lots of interesting discussion and things to think about!

This issue will be general notes and things I've seen this week that I wanted to start collecting. I'll do a more formal and organized writeup in another issue after the week is over.

[joehand]

Research Data Alliance - Working Groups and Interest Groups of Interest

Research Data Alliance is an open collection of academics, data managers, etc. A good primer here. They meet regularly and focus on interest groups and working groups.

There are lots of interesting working groups and interest groups in the Research Data Alliance. A few already have specific recommendations out, that may overlap with Dat:

• Data citation working group recommendation
• Persistent Identifiers recommendation
• (more soon)

---

There are lots of working groups and interest groups that are in progress, i.e. formalizing their recommendations, that overlap with Dat. Here is a list of some. Someone presented a good doc that summarized the groups I'll try to get that and update this later with links to their docs:

• Research data provenance
• Repository platforms for research data
• Publishing data working group
• E-infrastructure
• PID ig
• National data services
• Long tail is research data (very relevant)
• Data description interoperability
• Research data repository interoperability
• Publishing data services

[joehand]

General Things of Interest

• DataFAIRport - Their vision is to join and support existing communities that try to realise and enable a situation where valuable scientific data is ‘FAIR’ in the sense of being Findable, Accessible, Interoperable and Reusable.
• Scholix (elsevier) - high level interoperability framework for exchanging information about the links between scholarly literature and data. It aims to build an open information ecosystem to understand systematically what data underpins literature and what literature references data.
• Maslow hierarchy for research data - This is really interesting to think about. Seems like Dat is focused more on the bottom two sections.
• Preserving data at risk

Data Repositories of Interest

• BioCaddie - A meta data repository ( collection of other data repositories). Recently released, they've done a ton of work on ingesting other data repositories.
• ProteinDataBank - 555GB of protein data. Specifically, there may be good opportunity for us to test large scale sharing and versioning. They currently have this versioning scheme and downloads over FTP.
• datasearch - elsevier - meta data repository
• share - open science framework - meta data repository

Data Versioning, PID

• Identification of Reproducible Subsets for Data Citation, Sharing and Re-Use

[joehand]

Totally Random Thoughts/Meeting Notes

Collecting random things here! woot. Slides that were good here, I'll try to link those to specific notes where relevant later.

Long Tail Data IG Meeting

• The interest group of long tail data is very relevant to Dat. They are drafting these recommendations.
• Amazon succeeded by providing access to the long tail of books. There is a huge gap is services/infrastructure for the long tail of data. This long tail is from individual researchers. Generally small data (>1GB). 90% of researchers keep data on local, lab, or institution servers/backups.
• Most data isn't highly structured or has standardized metadata. But this is where a lot of the innovation comes from. Newer fields, newer types of data, etc.
• "Important to show immediate value upon deposit". "Needs to link to other resources". "Data that doesn't lead to publication is the data that gets lost".
• Big data, little data, no data - Book that may be of interest.
• Orphan data are the bigger problem, data that don't have a home.
• Libraries will provide for a key partner in managing and publishing this type of data.
• Metadata is important but need to strike balance between ease of use and standardization/discovery.

Joint meeting of IG Preservation e-Infrastructure, IG Reproducibility: Knowledge Preservation Tools for Researchers

• How can we foster relationships in overlap between researcher and archivist? Bridge the gap between researcher with data and preservation needs. What tools are necessary to make this easier?
• If you want to preserve it, you have to share it. Non-trivial thing and can be daunting for researchers.
• If you can make your workflows work on remote computers, you can preserve it for reproducible execution.
• "No generic preservation tool exists" (assertion). Must be some level of commonality to preserve all science & research inputs (data, software, workflow, etc.). Capture should complete for reproducibility.
• "Not tool will be adopted unless there is economic incentive (enlightened self-interest)"
• What is meant by reproducibility? Trying to get the same answer or independently implement the same methodology to see what answer if given. There are always people reproducing others' work. usually students, postdocs. No way to publish or credit them. We know they are happening But not what happens or who is doing them. We only hear about them when they differ significantly from the original result.
• Major risk using these tools is what happens if a tool get deprecated, loses funding, etc.

Data Versioning BOF

• Colocation with petascale computers and cloud intensive cloud.
• Not just an issue for big data, also small datasets.
• query subsetting data is important; spatial queries useful
• "it is hard for a researcher to cite the exact data extract used to support investigation". particularly with subsets or dynamic queries
• Battle over who assigns the DOI/PID for dataset and subsets of data. If data is of value, institution or repository owner is prepared to store it. But who mints the DOI?
• Lots of different perspectives on how to version and what is relevant.
• What happens when storage is full and old versions need to be discarded?

Metadata, Standards, Interoperability

• Whoa! Tons of discussion across many groups of metadata, standards, and interoperability. This is a huge problem. There are many many (way more than I though) specific domain or more generalized data repositories. Some of these have common metadata (usually in domain specific case) but lots of data repositories have their own and rely on translations to/from other metadata formats in order to improve interoperability.
• Metadata is great in that it allows you to do things like improved visualization, validation, discovery.
• Data curation (making sure a repository has only high-quality data) relies very heavily on metadata.

Data Repositories

• Session on Sustainable Business Models for Data Repositories - This session covered the costs, budgets, and revenue streams for various data repositories. There are many different cost structures and was eye opening.
• Many groups trying to figure out the perfect metadata so they can improve data discovery and curation. Some are supposed to be more general repositories and some domain specific.
• There are over 1,500 data repositories.