Record comparisons

[allysonlister]

This is the start of a discussion about how record comparisons could be implemented within FAIRsharing, as there are a lot of factors to consider, and making some kind of mockup of how they might be displayed (e.g. what is in Allyson's head) is probably what should happen after I hear back from everyone about what is most likely and what is least likely.

Equally, this could be coded as part of the assistant, or as a separate interface to searching etc on the main site, would be good to know your views. At first glance, it sounds like something that would work well in the assistant.

What do we compare?

Certain metadata fields could be compared, e.g. subject/domain/taxonomy tags, the database conditions in databases, or the policy metadata. Others wouldn't be worthwhile to compare, e.g. description and homepage.

We could also find common resources that are linked via our relationships to each of the records being compared, e.g. displayed in the same style as a record graph. In this case would it only be immediate neighbours (which is the main utility I can think of), or would we allow e.g. up to 2 hops to capture those that might be close in terms of the resources they mutually relate to, but not a perfect match?

How many records do we compare?

At a minimum this would be a 2-record comparison, however ideally I'd like to be able to compare e.g. up to a dozen. Larger comparisons would very likely take completely different visualisations for some parts of records (e.g. metadata fields), but the same visualisations for others (e.g. points of commonality in relationship graphs)

For metadata fields that are being compared against e.g. a dozen records, we may wish to create something like a heatmap or similar summary to show where the highest incidence of matches of metadata is, e.g. to rank records' similarity.

[knirirr]

For this, there could well be various quite different goals for people wishing to compare records.
Before embarking on anything it's necessary to get a better idea of what they would be trying to acheive, i.e. what would they want to learn from such a comparison. Then, making a decision as to what to build will be possible.

[allysonlister]

I will work on a list of user scenarios around record comparison, added to our user scenarios document, which will hopefully help here. Once done I'll let you know here. Not sure yet how best to show this comparison, so will be great to hear from you (once the scenarios are complete) as to what kind of comparison would be good, e.g. textual, graphical, etc

[knirirr]

A comment by @allysonlister

The use cases for this remain very vague. Also, it’s probably another case where the (relatively) few users who want it all want different things. Therefore, I suggest delaying it for now.

[knirirr]

So far, the discussion appears to be revolving around the following:

• A REST API call which will take an array of record IDs and return some sort of JSON structure along the lines of "the different values for this field are X, Y, and Z and 3/15 records have X... etc. etc."
• The specific fields to be compared must be planned beforehand, taking into account any possible future features.
• Though the initial implementation will be REST API only, a GUI to submit records and a fancy visualisation of the JSON output may be added later.

[allysonlister]

Please see https://github.com/FAIRsharing/FAIRsharing-Strategy-and-Communication/issues/49 for a use case to have for an initial test of the comparison.