Feature Requested: "You May Also Like... this dataset"

[rebeccawilliams]

A la Amazon and other private sector recommendations:

Questions to answer:

• What tracking would need to be enabled to make this viable while respecting users' privacy?
• Is Google Analytics/DAP sufficient? CKAN analytics?
• Has a similar feature been used elsewhere in government or on open data catalogs?

cc @rypan

[dsmorgan77]

Possibly related is this problem statement from SpaceApps: https://2015.spaceappschallenge.org/challenge/data-treasure-hunting/

This doesn't necessarily require knowing about a user; data sets can be related to themselves. We have a folksonomy of relationships already through the topics and themes that we've been building on data.gov and the geoplatform also has communities and other controlled themes as part of the NSDI. We could use these manually curated relationships to seed a recommendation engine, find keywords that support the manual relationships, and unleash that on the body of existing metadata to see how it performs.

[nsinai]

Sounds promising! Is there a CKAN mailing list or group for the last question?

[rebeccawilliams]

@nsinai I don't know of another CKAN catalog implementing something like this currently. The CKAN extension for Google Analytics is pretty straightforward. See also #435 (it's pending the CKAN update)

@dsmorgan77 Good call. This approach could encourage more engaged keyword use.

2 examples of related data that come to mind (that are also relevant the Public Safety and are very newsworthy) are:

• UCR Hate Crime Data and the BJS’ National Crime Victimization Survey: https://fivethirtyeight.com/datalab/what-we-know-about-hate-crimes-in-south-carolina/?ex_cid=538twitter
• UCR's Justifiable Police Killings and BJS' Arrest-related homicides and CDC's WONDER database: http://fivethirtyeight.com/features/how-many-americans-the-police-kill-each-year/,

These examples would require the above respective databases to be better represented on Data.gov. (Cleaner metadata and as Collections).