Discussions on General Support

Is this something you plan to do on an ongoing basis? Our daily API calls limit is 100K.

Please remember to attribute back: http://www.freebase.com/policies/

Could you provide some more info about your site/project?

Thanks for using Freebase!

Hi Masouras,

Thanks for a swift reply! We are well aware of the attribution policies and we appreciate the reminder. The site / project we are working on involves incorporating information from Freebase into Silobreaker, a free free online search service for news and current affairs. Read more about it and try it out at http://www.silobreaker.com .

In the first step we would like to try importing images on a limited set of entities (400 000) using the API syntax described above. This is a one time task and we have implemented throttling mechanisms so that we don't violate the daily API call limits.

Is there anyone specific at Freebase that we can contact regarding collaboration and "joint ventures" in the future?

Best regards, Kristofer Soderholm Silobreaker

Hi Koffe,

Sounds like you've got everything under control re: the API limits, so you're good to go. One other thing I wanted to point out is that "fair use" images are limited to 150x150px (i.e. we won't give you anything larger via the API), which you've probably already discovered, but I did notice that your original post specified 500x500 so I thought I'd better mention it.

You can contact partners@metaweb.com re: collaboration/joint ventures.

K.

geojson to answer my own question

http://wiki.geojson.org/GeoJSON_draft_version_6#Geometries

is polygons up and running now?

any examples out there?

guess i need converters now

Sorry to interrupt the stream of consciousness flow...

Yes, the Geometry type is GeoJSON. I've requested that the owners of the type provide documentation (and posted a placeholder/my best guess in the discussion topics).

Like Wikipedia articles, the actual contents of the GeoJSON is stored in a separate content store which can be accessed by the 'trans' API (no, I have no idea what 'trans' stands for).

Working backwards from the final result, you have

blob http://www.freebase.com/api/trans/raw/guid/9202a8c04000641f800000000806d8ff

which is the same GUID as the corresponding Geometry topic on the MQL side of things:

http://www.freebase.com/edit/topic/guid/9202a8c04000641f800000000806d8ff explore view http://www.freebase.com/tools/explore/guid/9202a8c04000641f800000000806d8ff

which is linked to from the original Location topic at:

Topic 1937 Oakland Hills Fire http://www.freebase.com/edit/topic/en/1937_oakland_hills_fire

If you have your web browser set up with Google Earth as a viewer for KML files, you can use this URL http://www.freebase.com/api/service/geosearch?location=/en/1937_oakland_hills_fire&format=kml to fetch the shape in a form that you can view in Google Earth. The formatted KML also includes a link back to the original Freebase topic that you can use to navigate in the other direction.

Hope that's enough to get you started. If you add geo info for Ireland, I'd love to hear about it. I used to live in Galway and 3/4 of my ancestry is Irish (Roscommon, Clare, Kerry).

Hi Dublinstreams,

To add politicians to Freebase, go to Politician and click on the "Add More Topics" button (upper right). Enter each name, or else click on "Go to the import tool" to paste in a list. Once you've added them, you can edit each one to specify which governmental body they're part of. You should do this using the "Government positions held" property; the 30th Dail would be under "Legislative Sessions".

You could also start from 30th Dail and edit "Members" there. If you wanted to import a list, you could click on the little arrow next to the "Edit" button and choose "Import list".

okay, might do a bit more fiddling in sandbox to learn instead of here, now that i found it, you don't make it easy to find your help content... or basic start advice.

so will you pull in their constituency and positions and date of assumed office too? dob etc etc

I was wonder having connected to the wiki, why you didn't pull in the other info too...

the problem is that the infoboxes on wikipedia most poltician pages, they have "assumed office" for both the date they win an election and they date they are appointed to particular position in government or party, which is quite unclear... i was hoping to scrap the info for timeline, but even dbpedia uses termstart for both coming from wikipedia... as will you's i suspect...

have to say how long should i wait, often website promise something soon as years pass by

You're preaching to the choir as far as the "Getting started" docs go. Another good source of information which you may not be aware of is the archives for the 'developers' and 'data-modeling' lists. The wiki is new, so it's not well populated yet. The good news is - it's a wiki, so as you find stuff that you think should have been there, go ahead and add it. It's hard to replicate the viewpoint of a fresh set of eyes.

As far as timing goes, the developer who's working on the politicians already did a trial load on the sandbox (although it didn't include the TDs because the topic wasn't typed correctly yet), which usually means things are pretty close. I'd guess closer to days than months, but I'm not speaking for him.

The second page has new references but still misses the above:

http://www.freebase.com/view/en/lorenzo_ferrero/-/base/references

Michael,

Composer, Author, Dedicator are types from the Commons, these are NOT MY BASES, and the information that is missing now was there for donkey's years ! It only disappeared recently, since the "age 58 years" bit appeared a few days ago. I noticed that sometimes, when new information is added by the computer, other old information disappears. I think this is the case here too.

Here is a list of the missing things:

72 works composed: http://www.freebase.com/view/en/lorenzo_ferrero/-/music/composer/compositions

Record labels: http://www.freebase.com/view/en/lorenzo_ferrero/-/music/artist/label

Book contributions: http://www.freebase.com/view/en/lorenzo_ferrero/-/book/author/contributing_author_to

Works edited: http://www.freebase.com/edit/topic/en/lo_studio_dellorchestrazione

Works translated: http://www.freebase.com/edit/topic/guid/9202a8c04000641f800000000e8f98ca

The dedication: http://www.freebase.com/edit/topic/guid/9202a8c04000641f800000000bdfea91

By the way, in the case of "Album contributions," wouldn't it be nicer to show a list of the ALBUMS instead of the boring list of the roles ?

http://www.freebase.com/view/en/lorenzo_ferrero/-/music/artist/contribution

As you see, a lot is missing and none of it comes from my domains. This is a page which is number 2 worldwide on Google (number 1 on some other search engines), and it has to be perfect.

Thank you

We have a lot of IDs available depending on what sort of data you have.  Generally speaking, you can see the IDs available by going to the Explore view for any given topic, eg. http://www.freebase.com/tools/explore/en/blade_runner

 (You can get to the Explore view via the developer toolbar.)

You need to put the gender part under the student node, like this:

[{
  "type":     "/education/university",
  "ns0:type": "/education/educational_institution",
  "name":     null,
  "/education/educational_institution/students_graduates": [{
    "student": [{
      "name":          null,
      "/people/person/gender": null
    }]
  }]
}]

esamsoe,

You can query Freebase to return dated values for a large number of domains - most are saved as a dated integer, but it depends on the domain. For example, company revenue is stored as a dated money value, but military casualties are stored per conflict- the conflict implying the date information.

Focusing on the military casualties (I'm a fan of the Minard infographic), if the military losses were only incurred during battles we could query for all battles during Napoleon's Invasion of Russia and plot their dates and casualties. Unfortunately, most of the losses were through attrition (malnutrition, cold and disease) which occurred during the long march, which I do not believe Freebase currently has data or a schema for.

If you wanted to model the Minard Infographic, I would suggest creating a new base. Perhaps creating a Attrition base, with a event with attrition type with a time mediated casualties property?

Freebase is only as good as the data the people put into it and the quality and coverage of that varies widely by domain.

Someone recently upload (estimated) US HUD foreclosure data for 2008, but there is no set of general loan default data as far as I know. Current data (eg last 6 months) isn't as strong as things which are stable and historic, at least in my experience.

Be careful not to rely on BFG since it's an unannounced experimental service.

From my casual observations, it would appear that a large part (all?) of the "related topics" query is just topics which share the same type(s) as the given topic. It's probably constrained to topics which have associated images as well to make for a more interesting display, but other than that, I'm not sure there's a lot of "magic" in the query. It doesn't actually seem to generate topics which are a very close match.

Do they use something like Freebase Suggest to choose the topics? If you so, you could constrain the widget to only show types or domains, you could just show them new things in that type or domain. If you let them select a specific instance/topic, you've got a harder problem. It wouldn't be too hard to come up with a similarity metric that would give you a score for two topics, but it'd be a lot harder (impossible?) to get MQL to spit out a list of most similar topics. It's not really set up to do a n-space vector search like that.

Something you could experiment with would be to use reflection to get all properties and their values and then do queries based on that set of properties, relaxing the search criteria (in some unspecified manner) until you get a small, but non-zero, list of results.

You'd probably get more satisfactory results with less work, but just picking some popular cases and hard coding support for them directly. You could probably do Films, Movie Stars, Bands, Sports Teams, and a few other things and cover a good proportion of what people are interested in with relatively little effort compared to coming up with a general purpose algorithm.

http://www.freebase.com/history/view/user/still_alive

shows what he did to my bases and topics

You are correct I do apologize.

I was mad, sorry, I spent a lot of time making what I did just to have it whipped out over night.

I've deleted the post.

For what it's worth, the Freebase staff have the ability to perform a mass reversion of a user's edits. It's not an instanteous process, but it's certainly something that can be done in the case of vandalism like this.

OK I will try this again.

What about the user still_alive?

It's not hard to tell he just made the account to deface my base.

addition:

it seems like only one headquarter-address can be filled in?

f.

My guess is that the query to populate the next page is timing out. In my experience Freebase doesn't give you an error in this case.

The estimated number of entries in the table is nonsense for anything other than a straight type table with no filters (in which case it will be the number counted on the last nightly tallybot run). It should be ignored.

According to the details it provides it's a 500, which means it could be pretty much anything on the server side -- my guess is that there is some output on stderr or stdout that's being generated before the HTTP headers that is being logged. I'll file a bug in JIRA after I get home and repro a couple more times to give the admins something to look at.

Thank you for clarifying the count of entries. I'll play with ACRE a bit and see if I can figure out an efficient way to get the actual number of entries. I haven't played with the API yet, and this is as good a way to start as any.

Thank you,

is there also the possibility to use an operator like "OR", e.g. Country name is "United Kingdom" or "Ireland"?

best regards,

Frank

Opened a bug, CLI-9193

It's because the client is not accepting edits to boolean properties that are not set to restricted to a single value.

"model_years" is a non-unique property, but "null" can only return one value. (There's a quirk in MQL error messages -- I wouldn't call it a bug because it can sometimes be useful -- in that if you do a null query against a property with multiple values, it will return the values in the error message. This "result" can't be sorted, however.)

This query should do what you want:

[ { "type": "/automotive/model", "make": "Ford", "name": "Ford Explorer", "model_years": ["name":null, "sort":"name"] } ]

The query you are running has multiple model years, so you would need to ask MQL to return a list of them:

{ "type": "/automotive/model", "make": "Ford", "name": "Ford Explorer", "model_years": [{ "name": null, "sort": "name" }] }

or alternatively limit by 1:

"model_years" : { 'value' : null, 'limit' : 1 }

What type of network do you want to visualize? A lot is going to depend on the source of the data, the form it's in, the type of graph it represents, how you want to visualize it, etc, etc.

As one very specific example, if you wanted to visualize the graph of Freebase types, you could use something like http://schemaviz.freebaseapps.com/?domain=/book It uses GraphViz to create its visualizations which is a package that can be adapted to a lot of different uses, but there are also lots of other choices (including ones which are more interactive in nature).

There is a geosearch API, but there's nothing in the web client that uses it. The best that you can do through the Web interface is use the Location->contains property to constrain things. For example, here's a list of all National Register of Historic Places sites in NH http://www.freebase.com/view/user/tfmorris/default_domain/views/nh_national_register_of_historic_places.

Depending on how the data was populated for the particular type that you're looking at, you may be able to constrain by county (or not at all).

The geo API is an http API that can do shape and proximity queries using the freebase data. The freebase.com site uses that API to produce maps whenever location information is available. For example, here is a map I created that shows all the mexican restaurants that we have locations for in Freebase:

http://www.freebase.com/view/user/masouras/default_domain/views/mexican_restaurants

In the docs that I pointed you to, there are some examples of how to do proximity and shape searches. What exact query are you trying to produce ?

The geosearch API doesn't really speak MQL, except for its input and output filters. It's a separate API of its own. Don't feel bad about being confused, I just had a peek at the documentation and it definitely isn't focused on the needs of a newcomer.

Here's a query that you can use as a starting point:

http://www.freebase.com/api/service/geosearch?location=/en/suffolk_county_massachusetts&type=/base/usnris/nris_listing&inside=true&indent=1&mql_output=[{%22id%22:null,%22name%22:null,%22/base/usnris/nris_listing/significance_level%22:null}]

which returns all National Register of Historic Places sites which are in Suffolk County, Massachusetts.

Note that the topic used in the location parameter in this style of query must have a shape stored in Freebase to work. If it doesn't it will silently return no results without telling you why. For example if you use /en/boston_massachusetts in the above query instead of Suffolk county, you'll just get an empty list, despite the fact that there are plenty of listings in Boston. To work around this, use one of the other search forms such as a radius search or a bounding box search.

Hope this helps. Don't forget to send me a pointer to your whizzy new web site that uses all this fancy technology when you're done!

Have you seen the help page about editing a topic? There should be a save button right next to the input box, as perthis screenshot.

When entering data, you may have to wait a second or two until the list of possible matches is displayed. You need to click on a match, or create a new topic before the save button is activated.

Agreed; flagging a topic for deletion (which requires community votes/consensus) is quite different from deleting a view that you own (which you can do unilaterally).

K.

Webpage is itself a compound-value type (commonly abbreviated CVT) (combining a URI and a text description), and the UI doesn't support nested compound-value types. The workaround is a bit tricky, though. The first step is to make the expected type be URI, which is the type for simple links. The next problem you'll encounter is that the UI places all URI and Webpage properties in the sidebar in the "weblinks" box, but since you're working on a CVT, the weblinks box is hidden. There's a property hint that will make the URI display like a regular property instead of in the box, but you can't currently set that flag in the schema editor.

You can either write a query to do that yourself, if you feel like tackling MQL (we'll help you, don't worry), or just tell us which property you want, and someone should be able to set it.

Unfortunately there is no UI for this. However, you can do it directly through MQL. Here's an example - substitute your own property id at the obvious place:

http://tinyurl.com/l7gnqq

This is a link to sandbox so that you can experiment freely without fear of breaking anything.

The tidbit that you may be missing is that Types are really just advisory and an object's Properties don't necessarily need to match its types (although in most cases confusing things happen if they don't).

The properties of /freebase/property_hints of are designed to be used as 'naked' properties which get applied to the object without applying the corresponding type (as Michael's example shows).

I started a base for this, http://schemapatterns.freebase.com/, as there are a few recurring patterns in the taxonomic relationships.

Your example of education/field_of_study/subdisplines and /education/field_of_study/subdiscpline_of is a hierarchy pattern.

Hi Iain,

Actually your database goes beyond what I had in mind, which was actually somewhat fuzzy.

Now I see the purpose of such a database is to capture global properties of a topic/property graph in Freebase.

I have read this discussion, and I must say I have to ponder this a little bit.

What I'd like to have is precise mathematical definitions expressed in whatever suitable formalism amenable to formulating actual algorithm to automatically extract from Freebase actual instances of your schema patterns types.

Thanks a lot for your very informative reply.

precise mathematical definitions expressed in whatever suitable formalism amenable to formulating actual algorithm to automatically extract from Freebase actual instances of your schema patterns types

Do you mean MQL? If so, I've written a short MQL expression which will extract all instances of the hierarchy pattern http://tinyurl.com/nxsswo

Hmmm,

No, I didn't mean extracting data from your database, but rather populating your database from the raw Freebase schema.

You want something along the lines of this query

[{ "type": "/type/property", "id": null, "schema": null, "expected_type": null, "unique": false, "master_property": null }]​

where result.expected_type == result.schema. I don't think you can do this directly in MQL, but it's trivial to post process the results in in Python on your favorite language. I can send you a little Python program if it will help. You don't really need unique:false, but I think it'll be true for the cases you are interested in and will speed things up.

Ugh! Rewrapped all my code after I posted it. Let's try again...

[{
  "type":          "/type/property",
  "id": null,
  "schema":        null,
  "expected_type": null,
  "unique":        false,
  "master_property": null
}]

So from your above MQL query+post-processing I extract pairs of types and properties such that the type is so to speak both the source and the sink of the property.

Then, for each such property I can run the MQL query that will return me the directed graph of topic node related by that property, and instance of that type.

Now, If I find out that this graph is acyclic, I have a forest, ie a set of overlapping trees that give me my sought-after hierarchies. If not, I can collapse the strongly connected components of the graph in so-to-speak topic clusters and from the resulting acyclic graph I can then on proceed.

Bingo ! I'll have my hierarchies !

Sounds too good to be true, I'll have to check that for myself.

Now, generalizing a bit, I see that the Schema Patterns database aims at categorizing the structure of a topic-property graph, along with the structure of the homomorphic structure of the type-property graph.

I think that this generalization is interesting, but sounds a bit too far-fecthed for my humble purposes.

Thanks Tom and Iain for you enlightening answers.

Wikipedia updates are collected on a live basis, but they are just stored for a batch update process that runs periodically. Brian just mentioned that a new update process is being introduced which will increase the update frequency from every two weeks to every week. Theoretically, I don't think there should be anything much older than two weeks, but there can be misses in the reconciliation process or other issues. Perhaps you could give some examples and someone could have a look.

As far as I know there's nothing adaptive about the update process which tries to be more responsive for "hot" topics. The one exception that I'm aware of is that there's a special "death bot" that grabs recent Wikipedia death info and fills it in on Freebase. I think this runs pretty frequently (nightly?).

Tom

Hi Frank,

Client logic displays only types in the Commons on the Consumer Page of a topic (i.e. url has /view/ before topic id). All types, including user types, are displayed on the topic's Edit Page (substitute /view/ with /edit/topic/). For the Einstein page you quoted, this is:

http://jewlib.www.jewlib.freebase.com/edit/topic/en/albert_einstein

I think the reasoning behind this is that it's best to show only the reviewed, high quality types to new visitors.

Unfortunately that also hides good quality user types which haven't yet made it to the commons.

However, you can get the base types to be shown by appending /-/base/jewlib to the end of the url as so:

http://www.freebase.com/view/en/albert_einstein/-/base/jewlib