Schema on freebase

[Jack Alves]

Within the next few days I need to start modeling how bibJSON data will be represented on freebase. I will explore existing freebase schema then determine what new schema should be created for BKN related schema. Freebase has a wiki which includes a description of schema related to written works in general and specific types of scholarly works like dissertation and journal articles,

http://wiki.freebase.com/wiki/Entering_Data_for_All_Written_Works

I would be happy to have someone tell me specifically how BKN related data and bibJSON schema should be modeled on freebase. Suggestions and comments are welcome before, during, or after schema development. I will send progress updates as I proceed. My plan is to incrementally add schema as needed rather than map out everything we may want in the future.

Before diving in, my sense is that most schema we need is already on freebase. It would be useful to understand any "must-have" criteria for BKN datasets. For instance, a user should be able to easily extract a specific dataset like Math Genealogy or Author Claim. There are various ways to meet that criteria like creating views, publishing a set of queries, or to link each data record to a dataset specific type or namespace.

For extra credit I want to explore techniques to curate data on freebase. My ultimate goal is to establish data authorities so dataset owners can easily accept or reject edits by non-members. I know there are some limitations in freebase that make this goal challenging for some classes of schema properties.

Again, all comments are welcome.

[Jim Pitman]

Jack, re

"I would be happy to have someone tell me specifically how BKN related data and bibJSON schema should be modeled on freebase"

I dont think any of us has the time or inclination to do that. I think you have
touched on the key points already in your message, which is to respect the
ownership of various datasets  by BKN participants, and let me add *always* to link back to the original data source as well as a BKN curated version if it exists.  It may be that the BKN curated version is eventually a version on Freebase. But we are a long way from that now. We have BKN curated versions of things on multiple sites, Fred's BKN people site, my sites, the Harvard site, the AIM site, and we are experimenting with data integration and reconciliation.
I would like to experiment also with the workflow

my version -> Freebase version -> BKN People version

so I think just work quickly and agilely at embedding BKN data in Freebase, so there is always something for participants to see. Then we can figure out what works and what not. Certainly, adoption of well-designed Freebase schemas to save us trouble reinventing BKN schemas would be great.
I'd rather go that way than the reverse.
---Jim

[Benjamin Kalish]

Hi Jack,

I believe there are still several open issues regarding the BibJSON spec. These include what dataset level metadata is to be included and how it will be represented, various questions about specific attributes, many unresolved questions regarding the use of structured strings, and probably other issues of which I am not aware. It is not even clear to me that everyone involved agrees on exactly what the scope of BibJSON should be: how granular it should be, how big the bibliographic universe to which it applies should be, to what extent it should differentiate between purely descriptive data and data meant for access. I don't know how important any of this is to the freebase schema, but I thought it might be a good idea to bring up some of these issues again.

Benjamin Kalish