Discussions on spencermountain
Start a New Discussion
-
-
-
I'm not sure what the OpenCYC link is meant to indicate, but it doesn't seem correct for it to be tied to a Wikipedia disambiguation page (or this topic derived from it).
-
Adding spencermountain to the CC list...
-
ah! good find. its a dbpedia error - http://dbpedia.org/page/Michael_Gross this is the first error i've seen like it. - deleted.
-
-
-
Hey, just wondering where you got the data to load the myspace user names and what metrics you used on what to include.
thanks
-
hey zach, for each band, i just pulled links to myspace from their wikipedia page. accept it if there is only one link. i had planned on testing the username with some similarity or somehying, but i found the results very acuurate already. >99% manual this acre app im now working on grabbing others from musicbrainz withthis and discogs, but havent run it yet cheers
-
it is ambiguous as its just an external link, but since we know its a band, and know theres only one myspace reference, the inference is very tight. the errors i found ended up being nearly always band members, who are also typed as musical artist because they have solo projects, but not personal myspaces for these projects- like /en/noel galleger connecting to myspace.com/oasis, which is not ideal, but not totally false. other errors were just badly formed urls.
i wanted to pull mybrainz and discogs first, as they don't have this issue, but got held up with https://bugs.freebase.com/browse/ACRE-1520 , which is now fixed :)
i can't think of a better way to do this, as myspace doesnt store mbids (i dont think) and you cant guess it based on band name like lastfm, as there's no correllation - see http://www.freebase.com/view/base/myspace/views/self_referential_myspaces
cheers
-
-
-
Hi Spencer. It looks like there were at least several topics mistyped as Person with a gender of Female by you on Oct. 1 using the "Category writer." I'm not familiar with this app, but can you take a look at any of the other topics that got typed through this path to make sure they're OK? BTW, if these are somehow linked to Wikipedia categories, those are notoriously poor quality.
Here's the ones that I spotted, but there may be others:
Comet line Angels of Bataan Changjiao massacre Oflag IX-C Tungchow Mutiny Pumping Iron II: The Women Supersize She
-
-
-
Any comments on this question I posed?
http://www.freebase.com/discuss/threads/guid/9202a8c04000641f800000000c9c75fa
-
-
-
Hi spencermountain !
Sounds like a great project. Jamie Taylor, our Minister of Information, has done some work in this area -- for instance, check out the Web Ontology Base:
http://www.freebase.com/view/base/vocabulary/views/ontology
or, more specifically, the /base/vocabulary/ontology types on one of the topics in the base, like FOAF:
http://www.freebase.com/edit/topic/en/foaf
Unfortunately, Jamie is out this week and won't be able to make Hack Day. But I think he'd be interested in hearing about what you come up with!
Brian
-
:) ya, then cloned and improved by shawn!
it'd be great to connect things at the property level, whether manually or empirically. dbpedia and cyc have already linked some schema for example.
i think we could get some distance just comparing, (even visualy) the freebase and dbpedia ontologies.
see you soon,
-
Will there be someone at Hack Day who knows how the current RDF sameAs links to DBpedia are generated? Is it something simplistic like just assuming that two things with the same Wikipedia keys are equivalent or is there more to it than that?
Understanding what's there would be a big help in understanding how reliable what's been done so far is and figuring out how to extend it?
-
As mentioned above, Jamie won't be at Hack Day, but Colin Evans (one of the other semweb guys from the data team) will be, and can probably answer your questions in this regard.
-
Yes, just look me up - I might do a session on freebase and the semantic web if there's interest as well.
-
Hi Colin,
A session on Freebase and SemWeb would be very useful. I am part of an NSF funded project (http://bibkn.org) that has a need for sameAs assertions for researcher and publication identity and would like to understand how to use Freebase to improve our processes.
So please do such a session.
Thanks,
Nitin Borwankar
-
-
-
-
hey kurt, curious about the Genders From Powerset load, i've always thought powerset should be giving data back to freebase, since the two projects are so complementary. was it a screen-scrape? how'd you do it?
cheers
-
-
-
Spencer, I've been using your wikipedia parsing tools on your website. They're very useful (I typed about 700 topics using them today!) - thanks for making them available.
I was wondering if these apps can be put into javascript on Acre (I think they're php currently -right?), and made available on freebaseapps? That way I could poke around in the source code and tweak things!
-
iain, i love you.
yes, i'm crazy-working on getting these things acre-ized.
here's a start - http://wptofb.spencermountain.user.dev.freebaseapps.com/index
its suprising how little available there is (outside metaweb) that mashes acre:script + http://en.wikipedia.org/w/api.php
gonna reproduce them all. i'll keep you updated.
-
-
-
What could you use help with?
-
check out motivated event. if you want. its been a struggle to find data of this type, as its a pretty unique topic to map. go nuts really.
-
-
-
How about a type for "stolen item" - with reciprocal links back to Robbery? Stolen Item should probably have a "Stolen From" and a "Thief" property too.
-
(where "Stolen From" points to a Crime Victim, and "Thief" points to a new type, "Criminal" (not sure if it should be a "convicted criminal" or not?)
The schema probably needs to account for things that are stolen multiple times...
-
ahh great idea. done. also, i hereby knight you alecf admin of crime. i need helphelphelp here on this base, its hugehugehuge-
-
i think 'stolen from' works with 'crime victim' already, as robbery is cotyped with crime. maybe this could recip into some sort of general 'ownership' type. check out Frankfurt art theft which maps this with 'art owner'
-
-
-
Don't know if you saw it, but I already started a Dinosaur base. I've added you as an admin in case you want to use that one instead.
-
-
-
spencermountain, I'd noticed that you had been discussing an issue you had with modelling sentences and cvt within cvt. (Complex Value Type)
Perhaps criminal conviction should be a cvt, but the sentence should be an ordinary type. e.g. '10 years incarceration' is a seperate topic with a type of criminal sentence.
It's not elegant, but would work as a hack until there is a better way to show it in the UI.
-
-
-
-
I thought I would start a new discussion thread, to keep it seperate from the unrelated Dams.
I don't know much about criminal law, but it seems a massive and very complex area to model. I will try to help as much as I can with the schema.
You have made a good effort here - each crime modelled as a seperate topic is the best way of going about this. Your concern over large numbers of crime topics is probably nothing to worry about, but if it is a problem then only notable crimes should be listed against a criminal.
My main point is that there is big difference between someone being accused of a crime and being convicted. (and even then, judgements can be appealed and overturned)
My thoughts on this are as follows:
A person can be accused of a crime, a suspect - they might have done it, but they might also be entirely innocent so cannot be called a criminal. The accusation might be made by a police force, or by the media or joe publicbased on some wild conspiracy theory. These could be listed as an accuser.
I have decided on a criminal defendant type to note a suspect who went to trial. Here we can link to a topic about the legal case. (I noticed a legal case type already in skud's Law base - I'm not sure if it is suitable for criminal cases but have used it anyway). As people can be involved in multiple cases, I have used a CVT for each appearance in court.
If the defendant is convicted, only then can they be tagged as a convicted criminal. If an appeal later overturns this convicted status, the person is then detyped as a convicted criminal, but should still be typed as a criminal defendant (with a link to the appeal case noted in the court case property).
I have made a start, but it is a huge task!
-
this is really good iain,
i totally forgot people can be found innocent. ~duh.
one thing that feels awkward is how a convicted criminal is also listed as a crime suspect -like Kim Dae-han
i get how its part of the other types. though things would feel smoother in the convicted criminal type, and also the crime type if convicted criminals are no longer crime suspects - a suspect is someone who has not had a trial maybe. im not sure how this would play out.
I'd make an 'acquitted person' type for our innocent suspects. you know? but i get how other types are sort of pushing for seperating a person's criminal information into pre-verdict and post-verdict types. but It feels goofy, to say if they were convicted of arsonthat they were a suspect of arson.
the accuser type works really well with the witch trials but for 99% of crimes its really just 'the police' or whatever. i say we drop it, or shift it to like a gossip or 'mob rule' type maybe. i may misunderstand it.
wikipedia has little structured data on this subject. we are truly in the unknown.
-
convincted criminal and crime suspect feels a bit awkward for Kim Dae-han for sure. For other criminals such as Al Capone who was only convicted for tax evasion, and not for any of his other criminal involvement for which he can only ever be described as a suspect; it makes more sense.
My thought on the criminal suspect type was that it will always remain, no matter if the case ever goes to court, or the outcome of the court, or if it goes to an appeals court, or the outcome of the appeals court, or a re-trial, or a presidential pardon etc. etc.. Whatever happens after the accusation the person will always be noted as a suspect or former suspect.
For the accuser type I was thinking more on the lines of witnesses and informants, such as John Dean, who implicate others, Nixon, in a crime. Even listing the police is useful, as we can link to a particular state's police department, or the FBI or the CIA, Britain's Scotland Yard , French Gendarmerie etc. etc.
The acquitted person type sounds interesting - I look forward to seeing what you come up for it. The unknown makes it all the more fun!
-
i'd like to see al capone with type convicted criminal for tax evasion and type acquitted criminal for the valentines massacre crime. im gonna give this a try. apologies in advance if it is disastrous.
lets be strong then on what it means to accuse. when the cia accuses(lets maybe change the term) they have a warrant for arrest, but not the guy on the radio.
come to think of it, maybe we mean 'who arrested' (wpediaarrest) isntead of 'acused'. if someone's testimony is relevant in the arrest, or john dean, who was a witness in the trial (after the arrest) should be listed there as part of the investgation, or trial whatever.The only people whos accusation can lead to a criminal trial is a police dept etc, afaik.
-
hey so ive done a bunch of work, changed alot, put in a bunch of data.
what do you think?
what do you do in dubai?
-
-
-
spencermountain - do you know if there is an Air Crash type?
A new user, rollie08, was contributing to the American Airlines Flight 383 crash. I started the topic off by typing it as a disaster, but, I was looking for an air crash type and could not find one.
-
There is an airline accident type in the aviation commons.
-
