Discussions on Film

We need more awards. There isn't even a listing for the Oscars yet. I'll get to work on it soon.

I just typed a couple hundred film awards to help you out. I haven't found a clear pattern to get the film awarding organizations, though.

http://www.freebase.com/view?id=%239202a8c04000641f8000000004ff9516

By the way, you may want to install the wikipedia article greasemonkey script that displays a scrolling iframe on your page so you can copy data from the wikipedia article.

http://www.freebase.com/view/helptopic?id=%239202a8c04000641f8000000004949373

Hrm. Apparently you can't delete or edit your own posts? I'm in trouble.

Nope, there's no post editing yet. We've got a number of discussion features in the works, and I believe this is one of them.

This is directly analogous to the music album/release dichotomy. A DVD is a release of zero or more films and zero or more television episodes. I would encourage you to try prototyping the DVD type within your private domain and see what challenges and successes you encounter.

We've been thinking about how to represent events and films, and your suggestions (now showing in ..., user groups) are interesting ones. If there's enough demand, we could modify the film type to represent these. Alternately, you're welcome to create your own types, perhaps something along the lines of "cult film" with the property "fan groups", or "new release" with the property "cinemas showing this film."

Hi Jeanne,

It seems like spoilers are typically property values, meaning specific values within a particular a property.

This suggests that spoiler is a flag that can be on any piece of data.

Another approach would be to note when a property is valid on a timeline. We've suggested a facility for doing this in 'real life', where any property can have a date that is begins being true and a date when it ends (for example "Soviet Union is a country"). The problem is, we don't yet have that feature, and it's only planned for our reality's timeline, not fictional ones.

I like this idea though. I bet there is a way to make it work.

I like the timeline idea - but for the fictional world (films/books and their characters) it would be nice to have a timeline with simple discrete units. The timeline would be the series of films or books - and each property could be placed on the timeline either within or between each. Then developers could ask Freebase questions like "What are the properties of Luke Skywalker after Empire Strikes Back?"

Some properties would be valid even before the first Node (better name?) on the timeline - like his name, gender... and anything else that would be fine to see in a movie trailer or read on the back of a book.

Until we get the fabulous timeline world - we have a choice for implementing spoilers:
1) Simple (but less robust): Add the Spoiler property to the topic... list the properties with spoilers in them
2) More Intricate (but more powerful): Do the song and dance required to make a Spoilers compound value property, associate it with the Film (and Book) base types and then associate topics to the right film or book for which they have spoilers -- and now again list the properties on the topic that have spoiler data revealed in the associated film or book.

Option 2 is more like the timeline approach -- but needs work done with the keepers of the Film and Book types.

I was going to mention the same. Based on Robert's description, the timeline property appears to be quite a bit like the spoilers property, ie a CVT that connects the property and times during which the said property was valid (for whatever meaning valid would imply in that context).

However I think that the spoiler CVT is a bit more specialized not only in its function but also in its contents to warrant it's own little slice of freebase. One of the major differences I see is the cardinality of the spoiler property, for e.g. any given spoiler may connect a film/book with multiple properties of multiple topics, whereas the timeline property (as I understand it) seems to provide a single dimension (well two but for spoiler info, I am guessing only the start 'time' is valid) to the spoiler property.

Furthermore from a user's perspective, I think it will be easier to understand and add instances of many spoiler properties to Luke Skywalker's topic than to identify each spoiler property, figure out the timeline during which it was valid and then apply the relevant value to the one or both the timeline properties. At first glance using the timeline for this purpose appears to be a bit cumbersome from a querying point of view too, because as I understand it the client would need to get all the properties associated with Luke Skywalker and then then run multiple queries for timeline values which would then have to be applied to the current view based on the view's timeline.

We're a bit split on what to do with the 'unique' constraint. Although it seems sensible to mark those things that clearly should have only a single value, we often don't set it because it takes away the chance to show ambiguity.

Of course, we don't consistently do this -- "date of birth" on person is, in fact, set to be a unique property. I personally think it should be relaxed. Somebody writing an application would use the first value and ignore subsequent ones.

Is this really important data to have? Because having the URL listed in the website property and entitled with the text 'movie trailer' seems good enough at this time. One reason I'd be reluctant to have such a distinct property is that trailer links tend to become defunct within a year or so after film/dvd release (at best they tend to move alot).

Unless there's an external permanent trailer db out there (I'm not sure how long trailers stay in Apple's site).

Thanks for the comments/questions...These topics may better be served by taking this conversation to the data-modeling list.

The modeling of various written/story/based upon might be a necessary discussion as this involves more than just Film.

Synopsis coming from official sites would be of a restrictive copyright and probably not permissable for us to store (IMO).

Thanks. Any of the official publicity materials (publicity, not marketing materials) have been legally cleared for such use. Anyone can make use of the resources so long as they use them for informational purposes and not sold or used to manufacture merchandise. For web sites at most the lawyers contact people when they do not display the appropriate copyright on a photo, but that's even rare these days.

How much text are we talking about...The main thrust of Freebase is as a machine searchable location of links to data, not necessarily a database in of itself. That's why we have the keys for Wikipedia and MusicBrainz articles and only a 'blurb' of text to represent that a fuller and more complete text description exists and is reachable through the key. I beleive the same design constraints would be true for a synopsis.

 If it were limited to one reasonable length of a single sentence it might be possible for 33,000+ films...But maybe even that would consume too much in the way of our graph's resources? I don't know as of yet...Let me ponder this.

As for the writing creditation...We currently have film writer and film story contributor...When it's an adaptation (for any media against any media), you add the type Media Adaptation which gives you properties for "based upon" (so that covers I think all of the cases you listed earlier).

 As for executive producer versus Assoc. Exec. Producer versus whatever-else producer, we could add an enumerated list property like we have for special performance type for Production Credit. It could list Producer pecking order types like Executive, Associate, etc.

I have lots of official text that can be broken down into the schema if the schema exists for it. Production notes, Synopses, character bios, actor bios, crew bios, photos, etc. I wasn't aware Freebase was only trying to link to and not contain data, that doesn't make any sense to me since you can upload images and actually enter data and not just links. 

And what about the validity of the data? Does anyone actually care about the validity of the data? I was under the impression the subjective data imported from wikipedia and IMDB was placeholder to aid in building out the database and priming the schema so we could refine it more through these discussions. I have the studio handing me the docs that are used to create the credits that appear at the end of the movie. To me it would seem like Freebase would want to use those to develop a schema for film credits. Is there any value to Freebase in me entering official data or is Wikipedia data what you want?

As for Executive Producer the top credits would always include Executive Producer and Producer as separate billings, but associate producer is one of those ancillary crew credits that may not get special billing and appear like the gaffers and all the other crew.I would reccomend just adding Executive Producer because the problem was people were filling in Executive Producers and producers (Produced by)  and both executive and normal producers were getting lumped together under one credit which is not appropriate. For example Mark Cuban may get credited as an Executive Producer because he owns a production company but he's not a producer who actually worked on a film. The opposite of that would be someone who did work on a film getting the same credit as an executive producer. That has the same effect as your boss getting credit for all your work. And again, it depends on how valid you want your data to be. Maybe I'm nitpicking or kibitzing, the reason I bring all this up is because I have formal structured data and I am just trying to enter it but have a couple schema additions preventing me from doing so.  

All very interesting items of data you possess.

As I mentioned earlier, we are more interested in helping categorizing and linking datasets, not necessarily in becoming one in of itself. We should take this discussion up with the data-modeling community in the email list I gave a link to above.

 As the Film domain is pretty well populated with data and several third party and internal applications built upon the current schema/data structure of film, any changes to the producer or writing types would currently break our API, let alone those outside applications like Cinespin (we are in late-alpha stage of our development cycle and there will be hopefully soon in the future the abilty to more gracefully modify/enhance properties within a schema).

 We don't import any data from IMDb (they are a restricted, limited use copyrighted site), we will simply provide a link/key to the  IMDb's data on that film, actor, etc. We don't store the whole WIKIpedia article, we have a few lines as a description to help identify the topic, in particular for search/linking purposes (we are not trying to replace WIKI but make it that much more accesable and useable).

 If you have data that you think fit within the parameters of Freebase (synopsises, writing creditation types) you could try modeling it in your private domain and on sandbox and then discussing it on the data modeling list and we can see about making it a future part of the relevant domain.

 We really appreciate all the interest and activity, we just want to help guide you as the same time you may be helping to steer our development of Freebase into a better fusion of mutual interests.

That's a bug, all right. It's being tracked internally. I'm not award of any workarounds, short of using the API.

I forgot about data dumps.  http://download.freebase.com/datadumps/2008-03-28/browse/film/

I take it back, these have guids instead of compound values.

so it's been about a month now - still no solution to the export problem?

We currently don't support exporting a large set of data (approximately around a 1000) via the export to csv feature. You either must use our data dump as mikelove discovered above, or write your own MQL query.

 You can see what our filter view page uses by hitting F8 key to get the debug bar and clicking on the 'View transaction log' (its the client MQL query event you want).

 You can use that as a starting point (hint, modify the limit or eliminate it entirely).

Here's a starter query to get all the film titles currently in the system:

 [
  {
    "limit" : 40000,
    "name" : null,
    "type" : "/film/film"
  }
]

I assume total gross earnings would be appropriate? I guess it could be similar to the estimated budget. I would like to have a date for the earnings total so a better more accurate comparison of over a century of grosses can be done.

Or is it that more detail is desired?

US-centric: US Domestic & Foreign

By Region: Europe, Asia, North America, South America, Africa, Oceano-Pacific, Antartica

Other?

 

I think the box office is clearer than total gross earnings, because the latter could include DVD sales, rentals, TV rights, merchandising, etc. For example, Pirates of the Caribbean: At World's End domestically made almost as much in DVD sales as with ticket sales: $280 M compared to $310 M.

I guess the world box office would be good, or maybe US as it's the 40% of the world movie market, rest of the world, and world (US + rest of the world). I haven't yet come across a website that breaks it down further that that.

>I would like to have a date for the earnings total so a better more accurate comparison of over a century of grosses can be done.

Do you mean you would like to have the data adjusted for inflation? 

How about adding box office for the country of ORIGIN and world wide? For example, a german movie doesn't often even get to US theatres, their box office would therefor not be very interesting while the german would...

Yeah, that makes sense. So, Worldwide Box Office, and Local Box Office.

Johann, characters definitely do have occupations and you can certainly fill them out. There’s just a limit to how much information about the character is visible from the actor or film, especially since the same character may be played by multiple actors and appear in multiple works of fiction across multiple media. Just fill out the character’s name from the performance, then go visit it and flesh it out with occupation, gender, place of birth, and anything else you want to add.

Yes, I know I can specify the occupation there. But this doesn't completely answer my idea.

1. It would be nice to have the occupation on the moviesite too, next to the characters name as sometimes this is needed for knowing who this character is. Like finding out that Star X played the doc in the movie as "Mr Y" might not help you that easy.

2.t would be useful to view any titles of the person in a seperat column. Else the Characters ar created as "Dr. Jed Hill" instead of "Jed Hill" with "Dr." in another viewed column.