Discussions on Help Center

Yeah, what you're seeing there is called a "synthetic view".  These exist for any type in the system.  You can't edit and re-save a synthetic view, you can only re-save as another one.

 If you want a gallery, I'd recommend saving a second one, then going into the "All views" page and de-featuring the one you don't want.

Hi Bryan --

Interesting suggestion, but it doesn't seem to work.

Check out the explore page for the synthetic view.  I've got the "Gallery" property on the view, but it still shows up as a table.

Brian

try running this query:

[{
  "id": "/base/horsefacts/views/equine_anatomy",
  "/freebase/query_hints/preferred_view": {
    "id": "/freebase/view_type/gallery",
    "connect": "update"
  }
}]

regarding 2, we now have email notifcations for discussion post replies, posts to domains you administer, and posts to you profile.  in fact you will both be receiving an email notification for this post right now.

more notifications are coming soon.  you can update your subscriptions on your settings page.

Create the table that you want with the right filters and right columns displayed, then click on the little gear icon on the top and select "Download as a file."

Don't forget that the Terms of Service require attribution back to Metaweb/Freebase.

Not sure what you mean about the heirarchical vs relational, but the database isn't either really.  If you're talking about the information displayed on a topic page in Label: Value form, you can get all that to come out in table views as well with the appropriate amount fiddling about the the table settings (click Edit View and add additional types/columns).

For data that's missing, feel free to add it if you've got a good source that's license compatible. 

I believe you can create a view off the commons and save it anywhere you like, including to a base, and this will work.  I just tested this on Sandbox by creating a view of female authors from the Publishing commons, saving it to a Test Base, then going to my own person topic and typing myself as an Author.  When I went to the base and looked at that view, there I was.

 The "base topic" thing mostly comes into play if you are using the "Add Data" feature from the "Edit Base" interface.  We're making changes to this, so the annoyances related to "base topic" will hopefully go away in time.

Chris - I'll do that for the future, but I'd also like to note that all of your documentation refers to the sub-domain, not the hyphenated domain.

Kirrily - That sounds like the problem.  I guess I was inappropriately seduced by all the "Make a base!  Make a base!" exhortations.  All I want is a couple of new types and a handful of public views which operate on the standard data set.  I don't want to have to manually update the views when new data gets added to the system.  Is there a way to do that?  For testing purposes, I guess I can just change my upload program to co-type everything with the funky /base/usnris/topic type (at the expense of another chunk of write quota), but that's not really what I'd want in the long run.

ps It's not good to exclude the /base/unsris/topic qualifier from the query if that's what's being used to generate the view.  It's needlessly confusing.

pps My original note should have compared the 4 vs 10 topics.  It's needlessly confusing omiting one of the values. :-)

Are you perhaps trying to do a GET rather than a POST with your servlet? A 405 is usually triggered by using an incorrect method for a Web service, and only POST is allowed for mqlWrite. Since your query works through the Query Editor, this is my first guess. You might also want to take a look at Chapter 6 of the MQL Reference Guide, which covers the mqlWrite service:

http://mql.freebaseapps.com/ch06.html

But im using the "doPost" method in a servlet created by Eclipse? Are there example MQLWrite java codes provided by Freebase?

Thanks

The fields in the "add more" menu represent a couple things. Ones that are filled in automatically are properties that were used to create the view. (The + symbol is for adding additional instances to that property, which is why you're getting two property values. It would make sense if you were adding something that was on the NHRP as both an NHL and a National Monument or something.) The other way properties appear in the menu is if they are marked as disambiguating properties. So the way to handle this particular case is to base the view off of the Natural or Cultural Site Listing CVT. 

However, in this specific case, I have a question. Based on the Wikipedia article, it seems like the National Register of Historic Places has five categories: building, structure, object, site, and district. Additional listings like National Historic Landmark, National Historic Site, and the like seem like they should be a separate listing, especially since some of these other listings should be treated as Protected Sites (i.e., National Battlefield Parks). Does that agree with your understanding of these lists?

Well, there are categories and then there are categories.  The National Register Information System (NRIS) schema calls that calls that a "resource type."  The things that I've been working with as a category in this sense are what the NRIS calls a "nominator name" and they also have a "level of significance" which can be International, National, State, or Local.

Here are the counts for some of the nominator names of sites which are of international or national significance (30 and 7119 sites respectively).

1723 National Historic Landmark
  26 National Park
  24 National Historical Park
  20 National Historic Site
  13 National Memorial
   9 National Recreation Area
   9 National Monument
   5 National Seashore
   4 National Battlefield
   3 National River
   3 National Capital Park

There are a couple dozen of these.  As you can see from the counts, the most common category, even for things which are supposedly significant is "none."

I've been mostly following the UK model for Listed Building, but this could be sliced a few different ways.  The U.S. National Register is a little unusual also in that although the National Park Service adminsters it, some things like National Memorials get designated by Congress and added automatically.

As far as protection status goes, the vast majority have no direct protection.  Even National Historic Landmarks have limited direct protection, particularly if privately owned.  Much of the protection is indirect (e.g. additional reviews required for federally funded projects).

I guess I still think of those nominator names as being discrete listings; certainly anything has as nominator value of National Park should be typed as a Protected Place with a "system" value of United States National Park System. Using "category" for these loses us the ability to easily distinguish districts from buildings from objects.

In terms of values like "level of significance" and "area of significance", it might be worth having a separate type for NHRP sites, since these sort of values seem to be specific to it; that would give us a place to put the enumeration property, as well.

I don't know what's going on with the cross-posting; I can't do it either. I'll file a bug on it.

Designing a new type would certainly make my head hurt a lot less, but I'm worried that it would weaken the ability to treat sites uniformly on a global basis.

Can't one distinguish buildings from districts by how they're typed?  I haven't done an exhaustive search of the NRIS database, but I'm not sure I'd put too much stock in the type.  I noticed for instance that they have all boats typed as "structures."

What do you think about the UK precedent of using "category" for level of significance?  Are they going about it the wrong way?

I can see arguments for both sides, but I think a UK Grade I Listed Building and a US National Historic Landmark should be treated similarly.

/common/uri_template/template is a unique property, so the second URI replaced the first.

 A workaround would be to create another property and template that property, however that may be a denormalization of your data.

I guess the client was confused then, because it let me add a second value to the property, reorder the values, etc.  It definitely didn't think it was unique.

If the topic page formatter thinks it's unique though, that's what counts.  I guess I'll have to figure out a way to provide a landing page or some other mechanism to work around the limitiation.

Try this: 

http://www.freebase.com/tools/queryeditor?q=[{%22id%22:null,%22name%22:null,%22t:type%22:%22/common/topic%22,%22type%22:[{%22id%22:null,%22key%22:{%22namespace%22:%22/common%22,%22optional%22:%22forbidden%22,%22value%22:%22topic%22},%22optional%22:%22forbidden%22}]}]&read=1

Neither would I! - just that I had asked a similar query a while ago and someone else had been kind enough to provide the above.

Be careful with attempting to type untyped topics, as I ended up creating a lot of bucket/catch-all types on subjects I know very little about - see my Politics, Philosophy and Economics base for example!

Also, be careful with meta topics - they are named 'x of y'; such as 'Economy of Angola' or 'Politics of Angola' - and cannot really be typed that well, (except perhaps as a book subject or field of study, but that isn't really too useful).  For these I think it is best to extract the facts from them (e.g. the GDP from Economy, or a list of politicians from Politics) and then delete the meta topic.  You can then fill in the 'Angola' topic with these facts.   

Typewriter looks cool, but it's pretty well hidden.  How do users find out about it?

Where does feedback on it go?  If you're not logged in it just gives you a blank screen when you click on a type without any hint as to what's wrong.

It also doesn't seem to be working off the live database since many of the "place of interment" listings already got typed as part of my Cemetery sweep.  If you click through to the topic from the list, you can see that they've got both Cemetery and Place of interment assigned.  How often do the lists get regenerated?

Typewriter is new in the last few days.  It's still in shake-out phase, but I'll be blogging about it shortly.  Meanwhile, you can see new Freebase apps by going to our app editor and checking either Recent Apps or Featured Apps.  (I should really ask Jason to feature Typewriter, not to mention putting an attribution on it linking back to Acre!)

 The source of the assertions is a database we have of "partial evidence".  Al tells me that the data is currently a couple of weeks old, but there's a plan to have it update nightly.  I could also ask Jason to have the Typewriter app itself skip any that already have been asserted.  Maybe he'll join this conversation -- I'll point him over here.

Others who are doing data cleanup tasks which can be driven off topic name keywords and don't want to be constrained by what Typewriter thinks is important might be interested in my Untyped application.  Give it a keyword and it'll find all topics which have that word in the topic name, but which haven't been typed.

Want to make sure that all things "Boston" are typed in some fashion?  This is your tool.

This should be possible to do directly from the topic table view, but it doesn't appear to be.

Thanks, tfmorris!  This will definitely be helpful for people who do that sort of thing -- me included!

The ability to specify a search term of your own for Typewriter is one that's been asked for already. I believe they're working on it.  The hard bit (oddly enough) is remembering what people have voted "no" on.

oh, and am I the only one who can't get Typewriter to work??  I can finally log in, but now I keep getting an error when I click on a type:

Unhandled exception: TypeError: Cannot read property "score" from undefined (/freebase/apps/hosts/com/freebaseapps/typewriter/queue#9)

• /freebase/apps/hosts/com/freebaseapps/typewriter/queue:9
• mjt/src/template.js:246
• mjt/src/linker.js:414
• mjt/src/linker.js:368

@evening --  I've commented that part of the app out (was for a future feature anyway).   But I'm guessing that you're still not really logged in.

Can you try one more time after hitting the browser refresh button with the shift key held down?

The next step I guess would be to actually look at your cookies...

Untyped now shows article blurbs along with topic names.  I thought of adding write support to allow you to type things from the index page, but for right now I'm going to force people to go to the topic page (which will hopefully encourage them to extract more structured data from the article).  I cleaned up a few other minor things and added some more query examples, but that's probably it for the time being unless people have requests.

As far as Typewriter stats goes, you could definitely go nuts with this, doing stats by user, by user by domain, by Wikipedia category (which appears to be what the source lists are), flagging low quality taggers, controversial topics, etc.

I tried the same thing you were trying, and got the same results.  I would suggest filing a client bug at bugs.freebase.com (if you don't mind).

Currently cemetery doesn't have any included types - we can add that.  When you type something in the client, the client handles adding the included types.  If you are going to type in MQL, you will have to make sure to add the included types in your type assertion.

The bug is https://bugs.freebase.com/browse/CLI-7404

Any suggestions on workarounds to get the desired list?

Thanks Jeff.

I just realized that my original query is actually *not* doing what I want.  It is returning all topics which have "cemetery" in their name and is not excluding those which are of type "/location/cemetery" as I'd hoped.

Do I have the nested query wrong?  How do I exclude things already typed as cemetery?

Never mind, I figured it out.

    "type" : [{"optional" : "forbidden", "type" : "/location/cemetery"}]

needs to be

    "type" : [{"optional" : "forbidden", "id" : "/location/cemetery"}]

Hi, Liyang, and welcome. Roger Federer isn’t represented by a URI in Freebase, but by a node in our graph database. If you add him as a topic to your Swiss People base (let’s say), you are adding the same concept, and any additional information you add about him will accrete to the original Roger Federer topic.

The way you know that someone has already described him is that if you start to create a new topic called “Roger Federer,” the Web client auto-complete box will show you that there is a topic of that name which is a Person, Tennis Player, Olympic athlete, Pro Athlete, etc. If you are looking for that Roger Federer, you would select it instead of creating a new one. On the other hand, if you were trying to describe your cousin Roger Federer the couch potato, you would create a new Person of that name instead.

Using the API, it is very easy to create redundant new topics, which is why we ask that API users be careful to reconcile topics before importing data.

Thanks Crism for quick response.

So it is fair to say that it is up to the GUI side (client auto-complete box) to hint us that an existing instance could be the one we want to describe, in our example, this pre-existing instance is called "Roger Federer", and we indicate that this instance is indeed the one we want to add extra facts to by selecting it.

However, in real world, different word could mean the exact same instance (topic). For example, some one has already described a camera topic (instance) call "Nikon D70" (a digital camera) and without knowing this, I start to add a new topic int the same camera bass, and I named it "D-70 by Nikon", in this case, can the auto complete-box show me the pre-existing "Nikon D70" instance - the point I am trying to make is that I am a little worried that there might be cases where we created multiple instances for exactly the same "thing" in the real world, and data aggregation will then cannot be done correctly.

The goal of the Web UI is to make reconciliation as easy as possible. The name is not the only thing that is used; the relevance algorithm is constantly being tweaked to try for the best possible results.

There will inevitably be duplicates created. That is why we have the “flag for merge” operation; when users spot redundancies, they can flag them for other users to vote on. If the vote passes, the topics are brought back together.

Hi Liyang, I won't repeat what crism has well explained, but I wanted to highlight one feature for noting the multitude of names for a topic: every topic has an alias field, or "also known as", under the topic name. If you come across a topic with other known (English) names, please do fill them in.

For example, "Gordon Sumner" is an alias (his real) for Sting; "American Automobile Association" is an alias for AAA. This simple feature has done wonders for effectively cutting down duplicate creation.

Your query is malformed.  Try:

{
  "guid" : "#9202a8c04000641f8000000003c94278",
  "type" : "/type/domain",
  "types" : [
    {
      "guid" : null,
      "name" : null
    }
  ]
}

Thanks you, cheunger

It's working now.

You're right, there is a lot of overlap between Freebase terminology and SemWeb terminology but I'm not convinced that Freebase would be any more accessible or easy to use if it adopted standard SemWeb terminology. As I see it, Freebase types and properties less sophisticated than classes and attributes in something like OWL Lite so to give them the same name might give people false expectations about how they behave. For example, you mentioned that Freebase topics are like ontology instances but in fact Freebase has a lower level type called Object which is like an ontology instance and Topics are more like SKOS Concepts (at least in my opinion). For the average Freebase member who just wants to make a base about their favorite sports team, these type of distinctions just aren't relevant. So I think that the folks at Freebase have taken the right approach in choosing to define their own set of terms that appeal to a much wider audience.

All that being said, the underlying technology that powers Freebase is very compatible with the ideas of the Semantic Web and there are efforts underway from people like Jamie Taylor to bridge the gaps between the two communities. In October, Freebase started publishing all of their content as linked open data. Shortly after that, Chis Bizer from DBpedia created a set of links from DBpedia instances to Freebase topics and just recently the Open Calais project has announced that their upcoming release will automatically annotate text with Freebase topic URIs. I've also been working on developing some of my own mappings between Freebase types and some common ontology types like FOAF in my Ontologies base. The next step for me would be to combine these two approaches to make Freebase data available as RDF data using common ontology types.

B.B., I think it might help if you went up a level and explained what your goal is in broader terms. It’s hard to understand how to help you from that description.

You’re talking about real topics, domain topics, etc.; in Freebase, we just call those topics, domains, types, and properties. And we don’t say, “Obama is a child of topic person,” but, “Obama is an instance of type Person.” We also already have types for Person and Animal; the other three seem interesting, but as with other type proposals, the big question is this: What properties distinguish these types and make them interesting?

Once you have created a type, you can add multiple topics to that type by using the view associated with that type. Take a look at any type in Freebase currently; you will see controls on that page for adding new instances. You can also use MQL writes; any operation that the Web client does is done via MQL, and can always be executed directly.

Please remember to do your testing on sandbox-freebase.com. (-:

Hi crism,

Thank you for your answer.

Freebase is constituted by a huge collection of topics and others, in general they are broad but we can say they belong to and are categorized into 5 groups as I metioned before in order to help users could understand and use them easily.

If I create type Abstract  (Action or Meterial), I'll care about its properties not deeply because I only need to focus on what topics'll add it as a type, for example, when say about "software, virus, expressionism, etc" people'll understand they are abstract definition not other. With type Action, if we have verbs: "repair, burn, compare, etc" we'll know that they have action. I'm concentrating on what topics adding other types.

Thanks for supporting

Every Object in Freebase has an ID and a name and as far as I know all the data commons and user-generated types are objects. Compound Value Types are objects so they have IDs and names but most people don't use the name property since its not displayed in the UI.

Primitive data types like Date/Time and URI are not objects so they don't have names or IDs.

This is description i found on download page:

Link Export - A full dump of Freebase assertions in a simple utf8 text format. This is a complete "low level" dump of data which is suitable for post processing into RDF or XML datasets.

So i beleive it is suitable for coversion into RDF. I have also seen some threads on different discussion boards where people has done it but i couldn't find any direct reccomendation on how to do it.

I forgot about the full link export. Freebase is built as a collection of links just as RDF datastores are built as a collection triples. So the link export is like a complete copy of the underlying Freebase graph.

I don't have any experience converting between Freebase links and RDF triples but you can get a good idea of how they map to one another by comparing the sample data provided on the download page with the RDF of the same topic as it appears in Freebase.

I'd also like to put in a shameless plug for my ontologies base which provides some mappings from Freebase types to common web ontologies. 

In Freebase, most of the topics are named entities like "software". Actions like "repair" are mostly represented in the schema as properties. If you'd like to model the relationship between "action" words and their corresponding Freebase properties I've created a simple type that can represent those relations for you. Once you've entered a bunch of those relations you should be able to start data mining some text. Another user was asking a similar question on the developer mailing list last week

Thinkbase is a really cool application that lets you browse through the relationships between Freebase topics. If you want to find those releationships in MQL you need to query the underlying links like this.

Thank you, narphorium.

It's great. I've just read yours. I'll try it immediately.

Hi narphorium,

I found it is ok, but I'm wondering :

-  How topics have their relationship in Metaweb?

- We could name these relationship, for example: "parent-child" in "human-Obama" and other relationships if it has?

I query by using MQL, but I understand it not clearly.

Topics are related to each other in MQL using links. Each link has a source a target and a property so in your example, the source is /en/barack_obama, the target is /people/person and the property is /type/object/type. This is what's commonly known as an is-a relationship but there are many different relations available in Freebase as you can see here.

Another common relation between topics is broader/narrower. This type of relation is used in several places in Freebase like Organism Classification, Location, File Format & Profession. I have started mapping some of these common relationships onto the SKOS ontology which is a popular framework for this type of work.

Thanks a lot, narphorium

I think Freebase is quite well for people discovering themself, open, large as its name. I'm asking myself:

- Does freebase evaluate the weight of a term? for instance, if I type "arnol", freebase suggests me "Arnold Schwarzenegger" in the first list. It means "Arnold Schwarzenegger" is more important than others?

Thanks you, narphorium

You support me very well, I'll try myself.

Some Freebase topics (like George W. Bush) use Wikipedia keys as their ID but there is no rule that says that they have to. By default, Freebase assigns the GUID as the ID.

Why is this breaking your website? Freebase-suggest should be giving you the correct ID (/guid/9202a8c04000641f80000000047953d8) when you handle the fb-select event. Then you just append it to http://www.freebase.com/view and you have a valid link to that topic.

I have a database that isn't ported to freebase. It was easier to link with the Freebase topics (like George W. Bush).

How do I do a nested query by id/name/keyword and return 1 guid,

take that guid and query its primary properties?

Reconciling an existing database with Freebase can be a challenging task  because Freebase has so much information that you often have several topics that share the same name. For example, Freebase has 179 different topics all named "London".

If you know the Freebase type that you're looking for you can use the Reconciliation Service to get a ranked list of cities named London, for example. In practice, a lot of the terms will have an obvious match but many of them will still have to be reconciled manually or by using additional heuristics.

Once, you've reconciled each topic in your database with its corresponding Freebase topic ID you can query the Read Service for the properties of an individual entry. There are no "primary properties" per se, just different views of a topic's properties depending on which type you're looking at.

I recommend reading the first two chapters of the MQL Reference Guide. to help famililiarize yourself with some of the basic concepts.

HI Narphorium,

Thanks very much, found it. Appreciate your help.

HI Narphorium,

Thanks very much, found it. Appreciate your help.

The MQL Cheatsheet covers this exact problem under Textual Comparisons.

In your case, I think you just need to add the following line:

"name~="  => "love",

awesome, thanks!

do you have any idea how to get this done, but with many different terms? the "|=" works to check for multiple terms, but it looks for an exact match, while the "~=" will only accept one parameter...

I don't think its possible to do that all in one query. You'll have to build a seperate query for each term and merge the results in your application.

Luckily Freebase allows you to send multiple queries at once. An example of how this is done is shown on page 2 of the cheatsheet.

Thanks cheunger, by testing your query, it run well

I tried by using query editor then it's ok, but in my C# program I set "#9202a8c04000641f80000000000ce52c" as the input parameter, and I got the url exactly as u seen above, however it didnt run.

Hi cheunger,

I did exactly what examples guide. I tested successfully with input parameter as "name", but it is failed with "guid".

private static string server = "http://www.freebase.com";
        private static string queryurl = "/api/service/mqlread";
        private static string trans = "api/trans/blurb";
string keyword;

string query = "{\"type\": [],\"id\": \"" + keyword + "\"}";
            string envelope = "{\"a0\":{\"query\":"+query+"}}";
            string url = server + queryurl + "?queries=" + System.Web.HttpUtility.UrlDecode(envelope);

            
            HttpWebRequest request = (HttpWebRequest)System.Net.WebRequest.Create(url);
            request.Method = "GET";
            request.ContentType = "application/x-www-form-urlencoded";
            request.CookieContainer = authCookies;
            HttpWebResponse response = (HttpWebResponse)request.GetResponse();
            StreamReader reader = new StreamReader(response.GetResponseStream());
            StreamWriter sw = new StreamWriter("freebase.txt", false, Encoding.UTF8);
            string content = string.Empty;
            content = reader.ReadToEnd();
            sw.Write(content);
            sw.Flush();
            sw.Close();

Hi,

I also used System.Web.HttpUtility.UrlEncode(envelope, but it's not different from UrlDecode() function.

The easiest way to debug this problem is to print your "url" variable to the console and enter it into a web browser. From there, the read service will return an error message which will help you figure out why the query is failing.

In this case, the problem is that your query envelope is constructed using the multi-query notation but you are passing it using the single query parameter. You need to use the parameter "queries" instead of "query".

Hi narphorium,

I used string url = server + queryurl + "?queries=" +", I aslo print the value of this url in debugger of C#, it's fine, but when I put it in the browser it failed. I'll check it carefully why. Thanks

If your first query was:

[ { "name" : "John Smith", "type" : "/people/person" } ]

Then all you have to do to get the id and guid is to add them to the query with null as the value and freebase will fill the values in.

]

The MQL Reference Guide goes over this in more detail. You should read though some of the examples to get familliar with how MQL queries are constructed.

Hi narphorium,

Thanks for your answer. I've read MQL guide, and did both two query seperately, but I found difficult to resolve it. I'll try more.

Thanks for your support, narphorim,

First, I have a query to get all parent types, for example:

{
"id":"/guid/9202a8c04000641f800000000001805e",
"type":[]
}

And the result is:

{
  "id" : "/guid/9202a8c04000641f800000000001805e",
  "type" : [
    "/common/topic",
    "/people/person",
    "/people/deceased_person",
    "/government/politician"
  ]
}

So I split it to obtain all types:"topic, person, deceased_person, politician".

Then did I make right query to retrieve "id" or "guid" of  these types by using this query?

such as:

You can combine two query in the right way to get exactly "id" of each type?

I did in this query:

{
  "id" : "/guid/9202a8c04000641f800000000001805e",
  "type" : [
    {
      "guid" : null,
      "id" : null,
      "name" : null
    }
  ]
}

It's ok if I use "guid" as an input, but suppose I use a name such as "People" like:

 {
  "name" : "People",
  "type" : [
    {
      "guid" : null,
      "id" : null,
      "name" : null
    }
  ]
}

It is not satisfied my desire.

If you type that query into the MQL Query Editor it returns the following error message (way at the bottom):

"Unique query may have at most one result. Got 100",

This is a common mistake that happens when you format you query like this {...} instead of like this [{...}]. The first form means that you are asking for a single object, the second form means you are asking for multiple objects. You can fix your query by enclosing it in square brackets like this:

] 

Thanks narphorium,

I've done and got that.

Hi Jean-Lou,

we've been having some technical difficulties with Google today and our technical support team is looking into it. I'll post back when we have more info, but you may want to try again later and see how things work out. If you still have problems, let me know what kind of error messages/etc. you get and I'll relay them to the appropriate people.

No, we do not explicitly block the Google proxy, and the issue that Phil was referring to is unrelated.

Could you give us more details about the exact request you were making, the exact error message, and the exact time?  (The time will help me locate anything relevant in our system logs.)

Hi there, 

you might want to check out this discussion on the developer's email discussion list:

 http://lists.freebase.com/pipermail/developers/2008-June/001753.html

The basic answer is: you can't because the description is stored in a file system rather than in the graph database, which is what MQL queries.  If you want to get the description you have to do it by means of the API trans service. Here are the details for that:

http://www.freebase.com/view/guid/9202a8c04000641f8000000006cf5605

You'll need the GUID of the content that you want to retrieve.

Thanks philg, but I think everyone likes retrieving description that is included in MQL query rather than using trans service or re-quering wikipedia. 

It's my opinion.

I agreed. Everyone wants the description. I get that it is painful to put that in the tree but geez that would have been a lot more convenient...

Don't take it wrong, but learning first how to use MQL ( ok, I get it, you don't like sql... ) and then learning that it is not what you have to use...

I bet you have that sort of topic every two weeks !!

Another feedback on the documentation :

Use a FAQ and a lot more examples of queries.

Examples are fast to understand. A lot more faster than everything else.

Just a page with 30 typical queries with description, image and stuff like that. That would be cool.

Useful, but hard to find !!!

I had no way to find these without knowing it.

It should at least be put on top

The Freebase Data Dumps are a good place to download all the topics of a specific type. From there, you can look up each GUI the using the trans service to get the description.

 Alternately, you can query the Freebase Read Service to get a list of topics and their GUIDs by type. For example, all people with articles. 

In addition to Ed’s suggestion… The API is HTTP-based, so any language with TCP capabilities can be used, which is to say, pretty much any language. You will find it easiest to use a language with libraries for HTTP connections and JSON serialization and parsing, but those are also thick upon the ground.

When you write information with the API, it is indeed the same as if the write had been done through the Freebase.com Web site. However, the Web client has some knowledge of the Freebase data model and generally keeps things clean; making direct writes via the API, it is possible to create broken shapes, so testing on the sandbox server is strongly encouraged.

If you are looking to load a lot of data, please contact us first; we have some internal prototype tools that can ease the process, and we can also help you test your code before unleashing it on the main server.

I would also strongly recommend joining the developers e-mail list, as well.

You should have a look at our Acre Server and App Editor:

http://www.freebase.com/view/guid/9202a8c04000641f80000000086ea59b

With app editor you can create hosted, server-side applications that use Freebase data. Better yet, you can clone files (like templates) and entire applications from other users, then tweak around with them to get them to perform in the way you want.

In a couple months we hope to have some more of the bite-sized example code you're looking for available, but, if you have some knowledge of HTML and JavaScript, App Editor should help you get started.

Which language would you like to use for your project? I'm sure there's someone here who could help you out with any of the mainstream programming langauges.

 If you know Javascript and you've got experience with something like ASP, JSP or PHP then you should be able to put together a small sample application in less than an hour by looking at the existing sample applications in Acre. I just started learning it last week and I found it really easy to pick up.