Kelley, this is great! Thanks. And since you already have so much  
written up, would you consider going a bit further and offering it to  
the code4lib journal? My reasons are selfish -- i'd like to be able to  
find and cite this in the future.

Later I may have a few comments.

kc

Quoting Kelley McGrath <[log in to unmask]>:

> We called it "FRBR-inspired" since it probably wouldn't pass muster  
> as an orthodox FRBR interpretation. We were looking to experiment  
> with a practical approach that we thought would make it much easier  
> for patrons to discover moving images in libraries and archives. If  
> you haven't read it, the "about" page gives a general overview of  
> our approach at http://blazing-sunset-24.heroku.com/page/about
>
> Our top level is a combination of FRBR work information and  
> information about what we are calling the "primary expression." We  
> haven't made any internal distinction between these two types of  
> information. This enables us to record together the data that we  
> think people expect to see about the generic moving image and  
> reflects the sort of information that is given in IMDb, the All  
> Movie Guide, and film and TV reference sources. This is also the  
> data that we would want to re-use in every MARC record for a  
> manifestation of a given movie.
>
> This also allowed us to get around some of the areas of more  
> orthodox FRBR modeling that we found unhelpful. For example, FRBR  
> doesn't allow language at the Work level, but we think it is  
> important to record the original language of a moving image at the  
> top level. In addition, RDA has mapped a number of functions, such  
> as art director, costume designer and performer, to the expression  
> level. We would prefer to present these at the top level. It is hard  
> to imagine a version of Gone With the Wind with a different costume  
> designer or cast that would still be the same work. So all the Seven  
> Samurai data you listed above belongs either to the work or the  
> primary expression.
>
> We mingle expression, manifestation and item information in the  
> version facets on the right. We don't show any explicit expression  
> records. In this demonstration we are not actually identifying any  
> unique expressions, although in the future we will probably want to  
> do this for what I think of as "named expressions." Since this is a  
> demo, we are working with a limited number of attributes and the  
> only expression-level facets we provide are soundtrack and subtitle  
> languages.
>
> In this sense, our approach is similar to the near manifestation  
> idea that Simon mentioned. We are not trying to assert that we have  
> identified particular expressions. Rather, we are trying to provide  
> a mechanism for the user to identify the set of items that meet  
> their needs. It is not clear to me that libraries are always in a  
> position to accurately identify expressions.
>
> Rather than providing a hierarchical view where the user selects a  
> work, then an expression, and so on, as is common in FRBR  
> presentations, we permit the user to begin at any FRBR level. The  
> user is invited to limit by as many characteristics as they desire  
> to delineate the set of things that they are interested in. They  
> only need to select as many attributes as are important to them and  
> no more. This may not meet the needs of all scholars, but we hope  
> that it will meet the vast majority of general purpose user needs.
>
> It's a bit of a different approach than I have seen elsewhere, but I  
> think it works particularly well for moving images. One of the main  
> reasons I think this is because of the types of expressions that  
> predominate in commercial moving images. I will try to explain some  
> of my thoughts on types of expressions below.
>
> 1. Expressions that can be reduced to controlled vocabulary options
>
> These are the most common types of commercial  moving image  
> expressions, especially in the DVD era. They are distinguished by  
> characteristics that such as
>
>   Soundtrack language(s)
>   Subtitle language(s)
>   Accessibility options (captioning, SDH, and audio description)
>   Aspect ratio (although in this era of widescreen TVs, full screen  
> modifications are less common)
>   Colorization
>   Soundtracks for silent films
>
> These can be full described based on standardized data (although for  
> the silent film soundtracks, this would involve multiple pieces of  
> information, i.e., musical work, composer, conductor, performer(s),  
> etc.)
>
> DVD often contain what essentially are multiple expressions in that  
> they offer multiple soundtrack and subtitle options and may offer  
> multiple aspect ratios. A silent film on DVD may come with alternate  
> soundtracks. All of these can be combined in various ways by the  
> viewer, which can make for a large number of expressions contained  
> in a single manifestation.
>
> 2. Named expressions
>
> These are versions that are different in moving image content due to  
> have been edited differently. Examples include
>
>   Theatrical release
>   Director's cut
>   Unrated version
>
> Although Martha Yee found a strong correlation between differences  
> in duration and the likelihood that two things represented two  
> different expressions, this doesn't always work. The archetypical  
> example of Blade Runner was released on DVD with five different  
> versions (http://en.wikipedia.org/wiki/Versions_of_Blade_Runner),  
> all of which had run times within a few minutes of each other. These  
> types of expressions would benefit from their own identifier and  
> some sort of separate display. In public and academic libraries,  
> this type of moving image expression is far less common than the  
> first type. There are no examples of this type of expression in our  
> sample data.
>
> Many more subtle expressions of this type cannot practically be  
> identified by the individual library cataloger because the  
> publishers do not provide the necessary information. Many films  
> released on DVD have been remastered or restored or modified in some  
> way, but it is not clear how to usefully or consistently record this  
> information even when it is provided in some form. For example, it  
> sometimes seems like every release of the Star Wars films must be  
> slightly different, but the videos don't come labeled in any way  
> that's useful for identifying them. There is a page at Wikipedia  
> tracking some changes  
> (http://en.wikipedia.org/wiki/List_of_changes_in_Star_Wars_re-releases) and  
> an enormous thread on the release of the original theatrical  
> versions  
> (http://sideshowcollectors.com/forums/showthread.php?t=12157).
>
>
> 3. Manifestations with additional content
>
> Many manifestation could be considered to be new expressions because  
> of the presence of additional content. These types of expression  
> don't affect the content of the moving image work itself. These  
> additions could be potentially treated in a couple ways and the  
> decisions of individual cataloging agencies are likely to vary.
>
> a) Additional content recognized as a work in its own right
>
> Any additional content is theoretically a work in its own right, but  
> there is a cost-benefit analysis involved in deciding to treat it  
> that way. In some cases, DVDs come with bonus features that contain  
> content that the library might potentially have bought (or has  
> already bought) independently. These would benefit from being  
> described as separate works. There are a couple examples of this in  
> our data set. If you do a search for Citizen Kane, you'll get the  
> movie plus a TV documentary called The Battle Over Citizen Kane.  
> Both of these have been issued separately, but the manifestation  
> listed as " DVD (2001)" under both titles represents the same  
> manifestation, which includes the TV documentary as supplementary  
> content. Whether it is necessary to inform users in some way that  
> these are on the same disc at this point or not, I am not sure.
>
> b) Undifferentiated additional content listed with the manifestation
>
> DVDs often come with an abundance of special features, most of which  
> are probably not worth the time it would take to describe them as  
> separate works. We have not included any of this type of information  
> in the demo, but one possibility would just be to list the content  
> with each manifestation.
>
>
> Merging the expression and manifestation facets gave us a simpler  
> interface and we don't think it harms most viewer's ability to find  
> what they want. The four levels of FRBR make a lot of sense from a  
> theoretical perspective (although it is easy to see that there often  
> are multiple layers of expressions and that works have many  
> recursive relationships). For moving images, in many cases, users  
> care more about the manifestation format (DVD vs. VHS vs. Blu-ray  
> vs. streaming) than about expression characteristics.
>
> There is also not always a hard and fast line between what goes in a  
> record as expression and manifestation information. For example,  
> Criterion Collection is generally recorded as a publisher. However,  
> for many users, it likely serves as a proxy for expression since  
> Criterion is known for the quality of its videos. According to their  
> website, "Every time we start work on a film, we track down the best  
> available film elements in the world, use state-of-the-art telecine  
> equipment and a select few colorists capable of meeting our rigorous  
> standards, then take time during the film-to-video digital transfer  
> to create the most pristine possible image and sound. Whenever  
> possible, we work with directors and cinematographers to ensure that  
> the look of our releases does justice to their intentions."  
> (http://www.criterion.com/about_us)
>
> Well, that was a bit of a long-winded reply and didn't really answer  
> your question, but I hope it was helpful in framing what we're  
> trying to do. This is still very much an experiment and there are a  
> number of data modeling problems that I glossed over in order to  
> make the demo work, but which would have to be resolved for a  
> larger-scale application.
>
> Kelley
>
>
> Karen Coyle wrote:
>>
>> Kelley,
>>
>> do you have somewhere documentation on which properties/attributes are
>> associated with each FRBR entity? I ask this in part out of my
>> ignorance of moving image cataloging, and therefore I am having
>> trouble translating from the FRBR documentation to what appears in
>> your prototype. I did my usual search on "seven samurai" and the
>> display (which I assume represents the Work) reads (in part):
>>
>> Alternate Title:
>>    Seven Samurai
>> Director:
>>    Kurosawa, Akira, 1910-1998
>> Genres:
>>    Feature; Fiction; Drama;
>> Language:
>>    Japanese
>> Country:
>>    Japan
>> Original Aspect:
>>    Full screen ( 1.37:1 )
>> Run Time:
>>    206
>> Color:
>>    B&W
>> Sound:
>>    Sound
>>
>> I'm curious as to which are Work attributes and which are Expression
>> attributes. Also, is there an example that shows one work and multiple
>> expressions?
>>
>> kc
>



-- 
Karen Coyle
[log in to unmask] http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet