Discussions on Education

Date formed/disbanded are good suggestions. As a unit of political geography, I'm not sure that unifying the organization and the physical location is wrong, actually -- we do the same for cities, countries, etc., which are treated both as geographic areas as well as legal entities.

I wonder if this should be a dated location, then, to capture the date properties?

I've a few examples of edge cases which aren't exactly school districts, and I'm not too sure how best to fit them in:

• alternative-education boards/organizations e.g. Waldorf/Steiner 
• School examination boards such as International Baccalaureate or the various GCSE boards.
• normally geographically confined school boards e.g. India's CBSE but which have expat schools e.g. Indian High School, Dubai under their affiliation.
• Companies with run private schools in various locations.

For Waldorf, Montessori, Reggio Emilia, et al., I'm not sure. They're really approaches or philosophies. Waldorf does have some agencies, but according to the WP article, they're more like accrediting agencies.

Looks like we could use an Examination Board type, doesn't it?

CBSE doesn't sound like a school district; it's hard for me to tell from the Wikipedia article exactly how to model what it does, though.

Great idea--I've emailed the Math Genealogy project folks to ask them if they have their dataset available for public download. This is the sort of thing that Freebase / Amazon scientific data ought to have, as they're probably struggling to deal with all the data integration / web issues at scale!

 Is there a base / site / community dedicated to helping bulk import data into Freebase in an intelligent fashion? Whether crowdsource or mixed initiative, perl / ruby or MQL + JS, I think this pipeline could really use some upgrades :)

Every time I visit their site I see "more $$$ for student data entry please," so perhaps they could even be convinced to switch over to crowdsourced data entry on Freebase.  Of course that potentially raises data quality, political control, etc issues.

Perhaps it's changed, but the last time I looked the bulk data entry story was one of a) email us your spreadsheet and we'll have our data entry folks and/or programmers do it when they get around to it or b) roll your own MQL.  A robust table importer could go a long way towards opening up the data import pipe.  Of course if the plumbers are still working furiously on getting the plumbing connected, perhaps big open pipes aren't a good thing yet :-)

Perhaps we'll see the same Wikipedia vs Brittanica debates play out with Freebase, who knows!

Big open pipes aren't always a good thing, but there's probably a set of users interested in building off of Freebase architecture/data that want to use (a) data that doesn't exist yet in Freebase or (b) data that exists elsewhere but isn't linked up appropriately yet.

This seems like a good area for Freebase to try and have an open source strategy, just an idea :)

I know that the OpenII initiative now has code up here: http://code.google.com/p/openii/updates/list but I can't find any documentation for this open information integration project just yet.

That would be useful, but could get complicated - for example St. Hilda's College, Oxford was a former women only college which now excepts men.  We would also need a property for 1st year men admitted.

Also if an instititute then decides to return to single sex status we would have a problem modelling that data.

Instead, I'd suggest a property 'gender inclusion', or something similarly named; which expects a complex value type, showing a combination of 'gender' with 'from' and 'to' dates.

Does it make sense to extend this idea to address the idea of any special group of people being permitted to attend an institution or join a group? For example, should it show when people of a specific race were allowed to attend a school or join a society or club?

Might be simpler to do with 2 different properties - one related to gender inclusion and one related to racial inclusion.

This opens a whole bag of worms, as there are a vast range of ways in which you can group humans - nationality, race, social class, social status, wealth, religion, postal districts and so forth.

Gender only has a few possibilities (e.g. single sex-male, single sex-female, or co-ed) and so is very well defined (with some other possibilities with transgender)

 However with other grouping criteria, such as Race, there may be hundreds of combinations - e.g. Race x and y are allowed, but not Race z, w or v..  Also, if an institute was all-inclusive would we then have to list all races? (or flip it about to list only excluded races). I think it very undefined and would be extremely tricky to model.

I would support a 'gender inclusion' property as it is fairly well definable, but would have to first see a draft schema in action for other properties.

I think that it is reasonable to keep this simple and go with Gender + Start Year + End Year. Additional details like grades would be VERY complicated since an educational institution could be K-12 or university or medical school or law school...etc.

I put up the property on the Educational Institution type, but I'm not sure that I did it the best way. You can see it in action on Vassar College. I just used the Gender type as the expected type for the sex accepted, but I don't know that it really displays what we want to see very well.

What if, instead of using Gender, we had a new enumerated list type with the values Male Only, Female Only, and Coeducational? That way we could assert that Vassar was Female Only from 1861 to 1969, and Coeducational from 1969 to the present.

I think either way works. We could call it "Gender Admitted".

If you can have any number of entries, then you could handle cases like Wesleyan University that was all male, then co-ed for a brief period, then all male again... and finally coed again. That would be 4 entries with start & end years, right?

I used "sexes" rather than "genders" because it's less ambiguous, but "admitted" is probably better than "accepted".

In the current model, Wesleyan would actually only require three entries: Male (starting whenever the school was founded and continuing to the present), Female (with a start and end date) and Female (with a start date, continuing to the present. (Like this).

Hmm. Campus seems to be used fairly loosely. In some cases (such as University of Michigan-Flint) it seems to designate a separate institution. In others (like the FIU campus linked above), it just describes disjoint areas occupied by a university.

In the latter case, it looks like it might best be described as a type of location.

For FIU Biscayne Bay, I get it being a Location, but then how do you link it to the University?  It is in a totally different part of Dade County than the main campus, so you can't use contained by or anything like that (to show the relationship).  Right now the only way I see to do this is flagging this campus as an Educational Institution and putting the parent as FIU, which isn't tehcnically correct (like you said, it is a location, not a separate institution).

Also, Biscayne Bay campus has some Schools that aren't on the main campus. Not sure if you'd want to get that detailed in the schema, but I can see a use to say Arts & Sciences are at these campus locations, while Hospitality is only at this particular campus.  Thinking out loud on this one while looking at their website.

Sticky problem indeed - I think Jeff is on the right track. 

I think the criteria might be whether the "place" grants their own degree.  If so its not a campus extension, but a stand-alone institution.  If they don't grant degrees then it is just another location of the institution.

My point is that if there are multiple locations of a university (aka campuses), there's no way to show that.

Unless we don't want separate campuses like this as topics, in which case we delete the Biscayne Bay Campus topic and put their address in the FIU topic.

I was right -- I had messed with this before: 

https://sandbox.freebase.com/view/education/accreditation

Let me know what you think.  I'll post this to data-modeling as well. 

I like the model.

In terms of data source within the US the Department of Education has a public-facing database that allows searches:

It doesn't support returning all results at once (too many, says the web site). Choosing the "WASC, Accrediting Commission for Schools" as the accrediting agency, however, I got 209 results back: not bad. Getting to the year of accreditation requires an additional mouse click, but I'm sure someone on the data team can figure out a more automatic way of getting all the data in one piece. Better yet, perhaps the Department of Education would be willing to share this data with the Freebase community in a more portable format.