Description of the SMAFE Film Analysis Engine used in the 'global'
section of milkbar.com.au from:
Adrian
Miles, InterMedia
University of Bergen and RMIT
(Abstract
from the Digital
Resources in the Humanities Conference, Edinburgh,
2002).'Searching
for Contexts: Cinema and Computing'.
Also
see how the SMAFE Engine has been used for Battleship
Potemkin
Digital
resources in Humanities Computing largely concentrate on critical
encoding of content for access and archiving. Such projects
often produce applied critical outcomes however, it appears
that such work is constrained in its applicability to other
humanities communities by either a concentration on textual
artefacts, or on its emphasis on the quantitative analysis of
data. This paper documents two recent film based digital resource
projects that offer low scale 'tactical' encoding with applied
research and learning outcomes, and models ways in which such
'middle level' resources may be relevant to a broader conception
of computing in the humanities.
The
SMAFE project (SMIL Meta Analysis Film Engine) is an initiative
that explores the viability of broadband networked film analysis
systems for the authoring and delivery of screen studies content.
In addition, novel applied research and learning methodologies
are being investigated within the development of the film engine
and its various iterations. Each project relies upon SMIL, custom
written CGI scripts utilising PERL and mySQL, and utilises a
QuickTime Streaming Server.
The
first project utilises the Odessa Steps sequence from Sergei
Eisenstein's _Battleship Potemkin_ (1925), one of the most famous
sequences in cinema history. The sequence was digitised and
then each shot encoded utilising a series of metadata categories
relating to film elements such as shot scale, camera direction,
screen direction, and composition. A database provides access
to this metadata so that queries can be made for any encoded
criteria, and search results list the series of shots that meet
the criteria. Users can then view each shot individually, view
the shot in context (we automatically roll back 10 seconds before
the shot and roll forward 10 seconds after), or elect to view
all the matching shots in sequence.
Eisenstein
has written extensively, and in considerable detail, about the
visual, thematic, and intellectual patterns of montage that
he utilised in sequences such as the Odessa Steps, and the metadata
utilised for encoding reflect this. As a result the system provides
a valuable research resource when used in conjunction with other
available material, including the sequence in its entirety,
the film, and Eisenstein's writing, as well as secondary commentaries
on the sequence, Eisenstein, and other canonical essays on film
montage.
However,
as a theorist such as Eisenstein argued, cinematic shots only
gain their particular significance because of the series they
are placed within - it is montage (for Eisenstein) that generates
meaning. As a result this project is limited in many ways as
it is little more than an quantitative analysis engine. It is
valuable when utilised in larger research contexts and it does
demonstrate the viability of the engine (keeping in mind that
other film content could as easily be encoded utilising other
metadata schemas), but its content remains fixed and in some
ways intractable - you search for close ups with left to right
movement, and it finds them and lets you view them. What the
shots mean individually is considerably less than what they
mean in their various sequences.
The
second project, "Searching" is based on John Ford's
1956 western _The Searchers_ and rather than encoding the film
around a particular set of cinematic metadata the project begins
from a hermeneutic claim, that "doorways in _The Searchers_
represent liminal zones between spaces that are qualities."
Doors, as they appear in the film, are encoded around a small
data set (camera is inside, outside, or between, and is looking
inside, outside, or between) and still images from the film
are provided. Here a search by a user yields all the stills
that meet the search criteria, and clicking on any still loads
the appropriate sequence from the film for viewing in its cinematic
context.
Unlike
the Odessa Steps engine this project becomes a much more open
analytical system. The sets of images that are the result of
any query require significant interpretation by the user by
contextualising them within the broad claims about doors and
liminality. To do this requires some familiarity with the film,
and leaves open the problem of what sorts of qualities the spaces
demarcated by the doors represent. In other words the engine
sets up the possibility for exploring the hermeneutic claim,
and whether or not it is answered by the evidence provided by
the engine is determined in the interpretive work required of
the user, not the system - the system simply provides the possibility
for exploring a thesis.
For
example, a search based on the camera being inside and looking
inside reveals that virtually every such shot in the film involves
one of two women, both of whom are very significant characters.
This pattern, which is surprisingly poetic and consistent, is
clearly significant but only becomes easily recognisable as
a pattern because of the manner in which the engine is designed.
What it means however, is not answered by the engine, it productively
generates new questions or hermeneutic claims.
This
makes "Searching" a 'discovery engine' so what is
developed in the engine is the process for the unveiling of
patterns of meaning, and this would seem to offer an engaged
middle ground for distributed humanities content that provides
access and critical activity.
These
projects are not large scale archival or critical encoding projects
but are a tactical appropriation of available media and computing
resources to achieve middle level research and learning outcomes.
They are desktop based and delivered, using existing proprietary,
open source, and W3C standards, and because of their small scale
nature do not require significant investments in design, building,
or maintenance. They are creative and critical interventions
into the domain of an engaged new media computing practice,
and their scalability and success is not in breadth of content
but in the processes of engagement that they offer. This engagement
is distributed as it is feasible, and relatively simple, to
produce numerous such engines - for example students could encode
_The Searchers_ around other sets of critical claims within
the same engine and an emergent 'hermeneutic engine' would rapidly
evolve. It is such 'middle level' solutions that enable a practice
based research culture to complement the project.
Adrian
Miles (2002)
Authored
by Craig BellamyŠ 1999, 2000, 2001, 2002, 2003
Last Updated :
