Publishing Data Workflows

[Guest post from Angus Whyte, Digital Curation Centre]

In the first week of March the 7th Plenary session of the Research Data Alliance got underway in Tokyo. Plenary sessions are the fulcrum of RDA activity, when its many Working Groups and Interest Groups try to get as much leverage as they can out of the previous 6 months of voluntary activity, which is usually coordinated through crackly conference calls.

The Digital Curation Centre (DCC) and others in Edinburgh contribute to a few of these groups, one being the Working Group (WG) on Publishing Data Workflows. Like all such groups it has a fixed time span and agreed deliverables. This WG completes its run at the Tokyo plenary, so there’s no better time to reflect on why DCC has been involved in it, how we’ve worked with others in Edinburgh and what outcomes it’s had.

DCC takes an active part in groups where we see a direct mutual benefit, for example by finding content for our guidance publications. In this case we have a How-to guide planned on ‘workflows for data preservation and publication’. The Publishing Data Workflows WG has taken some initial steps towards a reference model for data publishing, so it has been a great opportunity to track the emerging consensus on best practice, not to mention examples we can use.

One of those examples was close to hand, and DataShare’s workflow and checklist for deposit is identified in the report alongside workflows from other participating repositories and data centres. That report is now available on Zenodo. [1]

In our mini-case studies, the WG found no hard and fast boundaries between ‘data publishing’ and what any repository does when making data publicly accessible. It’s rather a question of how much additional linking and contextualisation is in place to increase data visibility, assure the data quality, and facilitate its reuse. Here’s the working definition we settled on in that report:

Research data publishing is the release of research data, associated metadata, accompanying documentation, and software code (in cases where the raw data have been processed or manipulated) for re-use and analysis in such a manner that they can be discovered on the Web and referred to in a unique and persistent way.

The ‘key components’ of data publishing are illustrated in this diagram produced by Claire C. Austin.

Data publishing components. Source: Claire C. Austin et al [1]

Data publishing components. Source: Claire C. Austin et al [1]

As the Figure implies, a variety of workflows are needed to build and join up the components. They include those ‘upstream’ around the data collection and analysis, ‘midstream’ workflows around data deposit, packaging and ingest to a repository, and ‘downstream’ to link to other systems. These downstream links could be to third-party preservation systems, publisher platforms, metadata harvesting and citation tracking systems.

The WG recently began some follow-up work to our report that looks ‘upstream’ to consider how the intent to publish data is changing research workflows. Links to third-party systems can also be relevant in these upstream workflows. It has long been an ambition of RDM to capture as much as possible of the metadata and context, as early and as easily as possible. That has been referred to variously as ‘sheer curation’ [2], and ‘publication at source [3]). So we gathered further examples, aiming to illustrate some of the ways that repositories are connecting with these upstream workflows.

Electronic lab notebooks (ELN) can offer one route towards fly-on-the-wall recording of the research process, so the collaboration between Research Space and University of Edinburgh is very relevant to the WG. As noted previously on these pages [4] ,[5], the RSpace ELN has been integrated with DataShare so researchers can deposit directly into it. So we appreciated the contribution Rory Macneil (Research Space) and Pauline Ward (UoE Data Library) made to describe that workflow, one of around half a dozen gathered at the end of the year.

The examples the WG collected each show how one or more of the recommendations in our report can be implemented. There are 5 of these short and to the point recommendations:

  1. Start small, building modular, open source and shareable components
  2. Implement core components of the reference model according to the needs of the stakeholder
  3. Follow standards that facilitate interoperability and permit extensions
  4. Facilitate data citation, e.g. through use of digital object PIDs, data/article linkages, researcher PIDs
  5. Document roles, workflows and services

The RSpace-DataShare integration example illustrates how institutions can follow these recommendations by collaborating with partners. RSpace is not open source, but the collaboration does use open standards that facilitate interoperability, namely METS and SWORD, to package up lab books and deposit them for open data sharing. DataShare facilitates data citation, and the workflows for depositing from RSpace are documented, based on DataShare’s existing checklist for depositors. The workflow integrating RSpace with DataShare is shown below:

RSpace-DataShare Workflows

RSpace-DataShare Workflows

For me one of the most interesting things about this example was learning about the delegation of trust to research groups that can result. If the DataShare curation team can identify an expert user who is planning a large number of data deposits over a period of time, and train them to apply DataShare’s curation standards themselves they would be given administrative rights over the relevant Collection in the database, and the curation step would be entrusted to them for the relevant Collection.

As more researchers take up the challenges of data sharing and reuse, institutional data repositories will need to make depositing as straightforward as they can. Delegating responsibilities and the tools to fulfil them has to be the way to go.

 

[1] Austin, C et al.. (2015). Key components of data publishing: Using current best practices to develop a reference model for data publishing. Available at: http://dx.doi.org/10.5281/zenodo.34542

[2] ‘Sheer Curation’ Wikipedia entry. Available at: https://en.wikipedia.org/wiki/Digital_curation#.22Sheer_curation.22

[3] Frey, J. et al (2015) Collection, Curation, Citation at Source: Publication@Source 10 Years On. International Journal of Digital Curation. 2015, Vol. 10, No. 2, pp. 1-11

http://doi:10.2218/ijdc.v10i2.377

[4] Macneil, R. (2014) Using an Electronic Lab Notebook to Deposit Data http://datablog.is.ed.ac.uk/2014/04/15/using-an-electronic-lab-notebook-to-deposit-data/

[5] Macdonald, S. and Macneil, R. Service Integration to Enhance Research Data Management: RSpace Electronic Laboratory Notebook Case Study International Journal of Digital Curation 2015, Vol. 10, No. 1, pp. 163-172. http://doi:10.2218/ijdc.v10i1.354

Angus Whyte is a Senior Institutional Support Officer at the Digital Curation Centre.

 

Share

MANTRA @ Melbourne

The aim of the Melbourne_MANTRA project was to review, adapt and pilot an online training program in research data management (RDM) for graduate researchers at the University of Melbourne. Based on the UK-developed and acclaimed MANTRA program, the project reviewed current UK content and assessed its suitability for the Australian and Melbourne research context. The project team adapted the original MANTRA modules and incorporated new content as required, in order to develop the refreshed Melbourne_MANTRA local version. Local expert reviewers ensured the localised content met institutional and funder requirements. Graduate researchers were recruited to complete the training program and contribute to the detailed evaluation of the content and associated resources.

The project delivered eight revised training modules, which were evaluated as part of the pilot via eight online surveys (one for each module) plus a final, summative evaluation survey. Overall, the Melbourne_MANTRA pilot training program was well received by participants. The content of the training modules generally gathered high scores, with low scores markedly sparse across all eight modules. The participants recognised that the content of the training program should be tailored to the institutional context, as opposed to providing general information and theory around the training topics. In its current form, the content of the modules only partly satisfies the requirements of our evaluators, who made valuable recommendations for further improving the training program.

In 2016, the University of Melbourne will revisit MANTRA with a view to implement evaluation feedback into the program; update the modules with new content, audiovisual materials and exercises; augment targeted delivery via the University’s LMS; and work towards incorporating Melbourne_MANTRA in induction and/or reference materials for new and current postgraduates and early career researchers.

The current version is available at: http://library.unimelb.edu.au/digitalscholarship/training_and_outreach/mantra2

Dr Leo Konstantelos
Manager, Digital Scholarship
Research | Research & Collections
Academic Services
University of Melbourne
Melbourne, Australia

Share

Research Data Alliance – report from the 6th Plenary

The Research Data Alliance or RDA is growing about as fast as the data all around us. It got off the ground in 2012 with the support of major research funders in Europe, the US and Australia and has since grown to over 3,000 members. The latest plenary in Paris set a new registration record of ~700 ‘data folk’ including data scientists, data managers, librarians and policy-makers. The theme was Enterprise Engagement with a focus on Research Data for Climate Change.

Not an ordinary conference

What sets RDA apart from other data-related organisations is not just the size of its gatherings, but its emphasis on making change. Parallel sessions are not filled with individual presentations of research papers, but of collaborative activities that lead to outputs that can be used in the real world. Working groups are approved by governance structures that coalesce around actual problems that cannot be solved by individual organisations but require new top-level approaches. They are required to produce their deliverables and close shop after an 18 month period. Interest groups are allowed to exist longer, but are encouraged to spin off working groups to address changes as they are identified through group discussion.

Hard-working groups

Since 2012, these working groups have produced some impressive deliverables and pilots that if implemented across the Web and across organisations and countries could speed up research and improve reproducibility. They are governed by an elected group of experts, worldwide. Some current active projects are:

  • Data Foundation and Terminology WG: defining harmonised terminology for diverse communities used to their own data ‘language’
  • Data Type Registries WG: building software to implement a DTR that can automatically match up unknown dataset ‘types’ with relevant services or applications (such as a viewer)
  • PID Information Types WG: Creating a single common API for delivering checksums from multiple persistent identifier service providers (DataCite and others)
  • Practical policy WG: building on a previous WG that collected various machine-actionable policies practiced by different data centres and repositories, this group will register the policies to move repository managers to move towards a harmonised set.
  • Scalable Dynamic Data Citation WG: to solve the difficulty of properly citing dynamic data sources, the recommended solution allows users to re-execute a query with the original time stamp and retrieve the original data or to obtain the current version of the data.
  • Data Description Registry Interoperability WG: to solve the problem of scattered datasets across repositories and data registries, the group build Research Data Switchboard linking datasets across platforms.
  • Metadata Standards Directory WG: By guiding researchers towards the metadata standards and tools relevant to their discipline, the directory drives up adoption of those standards, improving the chances of future researchers finding and using the data.

Members of the RDM team have been involved in library and repository-related interest groups and Birds of a Feather groups, where surveys of current practice have circulated.

Not all men at RDA! Dame Wendy Hall from the Web Science Institute leads a Women's Networking Breakfast

Not all men at RDA! Dame Wendy Hall from the Web Science Institute leads a Women’s Networking Breakfast – photo courtesy of @RDA_Europe

RDA and climate change

Climate science was prominent in the 6th RDA plenary. This was not only due to the imminent Paris-based United Nations COP talks, but indeed due to issues of critical importance for the world today. For some years, driven by the climate model inter-comparison work underpinning Intergovernmental Panel on Climate Change (IPCC) reports and the massive datasets from Earth observation climate science has been located at an intersection of high performance computing, big data management, and services to support and stimulate research, commerce, and governmental initiatives.

Assessment of the risks posed by climate change, and strategies for adaptation and mitigation sharpens the need to solve not only the technical problems of bringing together diverse data (social, soil, climate, land-use, commercial,…) but also to address the policy challenges, given the diverse organisations needing to cooperate. This is a domain that builds on services to give access to data, for computation close to data enabled by e-infrastructure (such as EGI), and one that requires ever stronger approaches to brokering these resources and services, to permit their orchestration and integration.

Among initiatives presented in the climate-related sessions were:

  • GEOSS – The GEOSS Common Infrastructure allows the user of Earth observations to access, search and use the data, information, tools and services available through the Global Earth Observation System of Systems
  • Global Agricultural Monitoring (GEOGLAM) initiative in response to the growing calls for improved agricultural information.
  • An RDS group focused on wheat – the volatility in prices, in part driven by climate unpredictability, has become a major concern.
  • The IPSL Mesocentre
  • IS-ENES developing services for climate modelling especially
  • Copernicus, seeking to “support policymakers, business, and citizens with improved environmental information. Copernicus integrates satellite and in-situ data with modeling to provide user-focused information services”
  • CLIPC will provide access to climate datasets, and software and information to assess indicators for climate impact.

Dr. Mike Mineter, School of GeoSciences and Robin Rice, EDINA and Data Library

 

 

Share

Data Visualisation with D3 workshop

Last week I attended the 4th HSS Digital Day of Ideas 2015. Amongst networking and some interesting presentations on the use of digital technologies in humanities research (the two presentations I attended focused on analysis and visualisation of historical records), I attended the hands-on `Data Visualisation with D3′ workshop run by Uta Hinrichs, which I thoroughly enjoyed.

The workshop was a crash course to start visualising data combining d3.js and leaflet.js libraries, with HTML, SVG, and CSS. For this, we needed to have installed a text editor (e.g. Notepad++, TextWrangler) and a server environment for local development (e.g. WAMP, MAMP). With the software installed beforehand, I was ready to script as soon as I got there. We were recommended to use Chrome (or Safari), for it seems to work best for JavaScript, and the developer tools it offers are pretty good.

First, we started with the basics of how the d3.js library and other JavaScript libraries, such as jquery or leaflet, are incorporated into basic HTML pages. D3 is an open source library developed by Mike Bostocks. All the ‘visualisation magic’ happens in the browser, which takes the HTML file and processes the scripts as displayed in the console. The data used in the visualisation is pulled into the console, thus you cannot hide the data.

For this visualisation (D3 Visual Elements), the browser uses the content of the HTML file to call the d3.js library and the data into the console. In this example, the HTML contains a bit of CSS and SVG (Scalable Vector Graphics) element with a d3.js script which pulls data from a CSV file containing the details: author and number of books. The visualisation displays the authors’ names and bars representing the number of books each author has written. The bars change colour and display the number of books when you hover over.

Visualising CSV data with D3 JavaScript library

The second visualisation we worked on was the combination of geo-referenced data and leaflet.js library. Here, we combine the d3.js and leaflet.js libraries to display geographic data from a CSV file. First we ensured the OpenStreetMap loaded, then pulled the CSV data in and last customised the map using a different map tile. We also added data points to the map and pop-up tags.

Visualising CSV data using leaflet JavaScript library

In this 2-hour workshop, Uta Hinrichs managed to give a flavour of the possibilities that JavaScript libraries offer and how ‘relatively easy’ it is to visualise data online.

Workshop links:

Other links:

Rocio von Jungenfeld
EDINA and Data Library

Share