Discussions on alanl

I'm very excited over this; I've been interested in topic maps for years but this is the first practical implementation!


1. Will transitive closures eventually work, ie. should I bother with filling out the "contains" field? Are there TM theoretical roadblocks to getting transitive closures to work or is it just a matter of coding?
2. Do you have any tools/tips for extracting information out of wikipedia topoboxes?
3. Are there any tips for reconciliation? I understand its one of the biggest headaches in TMs but surely there are some workarounds apart from clicking until my fingers get callouses? :-P
4. I think it'll a good idea to try to get as much Olympics stuff entered by august *and* to get powerset to accept the topics...whats involved in getting powerset to parse a particular type?

1) I should get someone else who understands it better to explain in full, but I believe it's a planned feature in the future.
2) Yes.
3) Are you using the list import tool? Have you read about our reconciliation API service? Check the blog.
4) We're working on Olympics schema right now, and have data loads planned. I believe powerset accept the topics kind of by magic; that is, an update to Freebase shows up automatically on Powerset without us doing anything. Someone at our user group meeting was saying that they have to set up a certain amount of grammar per *type* of thing (eg people, tv shows) but not on the topic level.

can you tell me about how to extract info out of wikipedia topoboxes?

I'm currently using the list import tool right now to get the hang of the range of types possible, and I'll get around to learning the MQL-write stuff soon. In the mean time why do some of the list imports allow me to populate multiple fields at a time (http://www.freebase.com/importer/list/en/melbourne_airport?propertyId=%2Faviation%2Fairport%2Frunway_information)
but not types?
(http://www.freebase.com/importer/list/aviation/icao_airline_designator?domain=)

OK, well, the wikipedia infoboxes thing is an internal tool at Freebase. It relies on a bunch of stuff we have at the office which isn't really ready to be exposed to the outside world. But I can tell you that we base our extractions on something like the WEX (Wikipedia Extract) dump that we provide on http://download.freebase.com/ -- so if you download that, you may find it easier to parse the infoboxes and use them for whatever you're doing.

The difference between the two importers you link to is that one is importing into a CVT (compound value type) while the other is importing a simple list.

Is what you want to do import a list of stuff and type them as multiple things simultaneously? Eg. here is a list of actors who are also authors. We don't currently allow that in any simple way, though you can paste the list in twice. Another option you might find interesting is this: take a look at any type page, do some filtering as you see fit, then click on "Use results" then "Add to another type". So by doing this you could eg. filter TV Actors to only show those who are on Canadian TV shows, then type them as "Canadian Actor" (if you happened to have created such a type previously).

Or to give a real example: go to type "Building", filter for "building type" = "museum", then add all the results to type "museum".

Hi alanl and welcome to freebase!

I just noticed this discussion where you were talking about some airport work you were possibly embarking on. I just wanted to let you know that I'm also working on some airport refactoring - namely, doing away with ICAO and IATA designators as first class topics and putting them directly on airports instead. If you're interested, check out this post to our data-model list: http://lists.freebase.com/pipermail/data-modeling/2008-June/000881.html . Information about joining the mailing lists can be found here: http://lists.freebase.com/mailman/listinfo .