Initial feedback user story 10 - grant outputs

[mfenner]

In this initial feedback I want to focus on the GraphQL used, and potential improvements.

• I would put the query into the readme
• I like the variable interpolation you do in the second notebook. I think that pattern facilitates reuse.
• the result is the same, but I would consider restructuring the query to be more explicit that we are looking at outputs for a particular funder. So instead of

works(query: "fundingReferences.awardNumber:777523", first: 75) {
    totalCount

use

funder(id: "https://doi.org/10.13039/501100000780") {
  name
  works(query: "fundingReferences.awardNumber:777523", first: 75) {
    totalCount

• use types { resourceType } to have a more specific description of the outputs, e.g. Presentation or Project deliverable.
• use dates { date dateType } to have more specific publication dates, enabling a bar graph that is by month and not by year
• use formattedCitation(style: "vancouver") instead of bibtex to format the metadata into a human readable format. This assumes that the notebook generates a nicely formatted document rather than a bibtex file to download (we could do that in another notebook).
• think about extracting the author information (many of them have ORCID IDs) and do something with it, e.g. a visualization linking them to the outputs. Affiliation identifiers are not yet supported by Zenodo.
• handle duplicates/versions, which is common with Zenodo. Happy to discuss how to do that automatically.

[datasome]

@mfenner, I have addressed all the issues - please see below:

Item	Status	Comment
I would put the query into the readme	Done	Included an example query and the user story
I would consider restructuring the query to be more explicit that we are looking at outputs for a particular funder	Done
use types { resourceType } to have a more specific description of the outputs	Done	See Type column in html table of outputs
use dates { date dateType } to have more specific publication dates, enabling a bar graph that is by month and not by year	Done	See Number of Outputs per Year bar graph
use formattedCitation(style: "vancouver") instead of bibtex to format the metadata into a human readable format.	Done	See the second html table. As suggested elsewhere, also included in this table Descriptions, and moved from table 1 to table 2 counts of citations, views, downloads.
extracting the author information (many of them have ORCID IDs) and do something with it, e.g. a visualization linking them to the outputs.+	Done	See the co-authorship chord plot at the bottom, as suggested elsewhere
handle duplicates/versions, which is common with Zenodo.	Done	As suggested elsewhere, checking versionOfCount > 0 as the flag to exclude a given output (because it's a version of another resource)

[mfenner]

@datasome Looks good. The chord diagram came out really nice. Maybe a few small changes will improve it:

• larger diameter to better see the connections
• smaller font size as some names are cut off
• consider sorting names by FREYA partner affiliation.

I think the bar chart outputs over time can be improved. I would include every month from project start December 2017 until project end November 2020, not just months with outputs.

[datasome]

@mfenner, I have done the following to address the above feedback:

• Included all months between Dec 2017 and Nov 2020 in the bar plot
• Enlarged the diameter of the Chord plot
• (Chord library does not currently allow for controlling the font size of the labels - to hopefully display the full names I have enlarged the plot area. If this does not prove sufficient, we could shorten the first names.

Please note that I have not sorted creator names by FREYA partner affiliation as GraphQL query does not return affiliations currently. Will revise once creator affiliations can be retrieved.

[datasome]

@mfenner, I have now addressed the following comments from today's meeting:

• Set resource type to 'General' if None retrieved
• Merged the two html tables into one:
  • Constructed formatted citation in code (with author names linking to their ORCID profiles); Removed Titles column - as the title is now seen distinctly within the formatted citation;
  • Removed Funders, Award Numbers, Award Titles columns;
  • Added Publication Date column;
  • Changed Descriptions to max 200 chars with ellipsis;
• Added a piechart of counts of works by resource type.

[datasome]

@mfenner, I've added a fix to display the chord plot in mybinder.org - following the work-around suggested in https://github.com/shahinrostami/chord/issues/1. This actually writes out the plot's html to the server and then displays that html inside the notebook. The fix also works in JupyterLab.

[datasome]

@mfenner, I have now documented in Markdown the notebook for user story 10. Please note that the top Markdown table appears borderless in Jupyter lab and mybinder.org, but somehow not in github.