-
Stavros Alifragkis (Athens) and Franois Penz (Cambridge)
Dziga Vertovs Man with the Movie Camera (SU 1929)Ontology-based
Metadata Annotation for Non-acted, Montage Moving Image Works about
Urban Cinematic Landscapes
1. Introduction
Th is paper wishes to introduce a consistent and systematic
method for analysing existing moving image montage-based works that
experiment with the cinematic iconography of urban landscapes. Th e
method has been modelled, using Dziga Ver-tovs Man with the Movie
Camera (elovek s kinoapparatom, SU 1929), as a concrete and
particular instance of an urban creative geography constructed by
montage-based sequences and scenes organised in themed episodes. Th
is research proposes a relatively fl exible and comprehensive tool
f or the formal and stylistic analysis of moving image works that
exploits and expands upon the categories introduced by David
Bordwell and Kristin Th ompson for the aesthetic appreciation of
movies in Film Art: An Introduction.1 Th e shot-by-shot stylistic
analysis of the mise-en-scne, the mise-en-cadre and the editing
techniques, coupled with an investigation of the formal structure
of Vertovs Man with the Movie Camera, aided mining and identi-fying
key concepts that could serve as the basis for domain ontology
construction. Ontologies, a shared and common understanding of some
domain that can be
1 David Bordwell/Kristin Th ompson, Film Art: An Introduction,
New York: McGraw-Hill 2006 [Orig. 1979].
-
Starvros Alifragkis/Franois Penz52communicated across people and
computers, explicitly describe an abstract model for a particular
phenomenon and identify the relevant concepts.2 In other words,
particular categories from Bordwell and Th ompsons infl uential
work, comple-mented with additional categories, could inform the
construction of an ontology designed to provide suffi ciently
detailed and extensive descriptions for a specifi c domain:
non-acted, montage-based moving image works about urban cinematic
landscapes. Th ese works have been generically referred to as city
symphonies, a genre that fl ourished in the 1920s and 1930s, and
even exists in various forms today. Th is rather fl uid term has
been used extensively to provide a sketchy description for what
constitutes an extremely diverse and loosely connected group of
mov-ies, which employ wide-ranging editing and cinematographic
stylistic approaches. Ontology-based annotation of montage-based
moving image works that appear to share a common interest in the
urban form, coupled with a statistical stylistic analysis and
interpretation of the accumulated metadata, could inform a
discussion towards identifying shared features in terms of family
resemblance rather than exhaustive defi nitions for city
symphonies. Th us, the shot-by-shot analysis of Man with the Movie
Camera constitutes a starting point where the shaping of the tool
takes place, a tool whose analytical potentialities will
continuously be sharpened as the research expands to include other
movies.
2. Formal and Stylistic Analysis
Th is research draws widely upon Bordwell and Th ompsons work on
reading, un-derstanding and appreciating fi lm form.3 Th e method
proposed here builds upon their basic distinction between formal
structure and style and aims, on the one hand, at exploring the
intricate set of relations that link stylistic techniques (a.
mise-en-cadre or cinematography; b. mise-en-scne or staging; and c.
editing) to the narrative form, and for Man with the Movie Camera,
on the other hand. Th ese formal categories for describing and
reappraising style have been comple-mented with additional features
borrowed from Vlada Petri and Graham Roberts analyses of the movie
and Jean Mitrys seminal work on rhythm in experimental cinema.4 Th
ey have been coupled with our own investigations on the nature
of
2 Rudi Studer/Richard Benjamins/Dieter Fensel, Knowledge
Engineering: Principles and Methods, in: Data and Knowledge
Engineering, 25, 12/1998, pp. 161197, here p. 184.
3 See Bordwell/Th ompson, Film Art.4 Vlada Petri, Constructivism
in Film: Th e Man with the Movie Camera A Cinematic Analysis,
Cam-
-
Dziga Vertovs man with the movie camera 53the urban landscape
depicted in the movie, based on research on the
urbanist-disurbanist dispute over the future of the socialist
city.5 Th e part of the analysis that focuses on style
mise-en-cadre and mise-en-scne in particular off ers itself to a
statistical analysis of quantitative and qualitative data. Th e
part of stylistic analysis that focuses on the foregrounding of
specifi c editing techniques combines statis-tics with the
detection of patterns to draw inferences. Th e ensuing exploration
of Man with the Movie Cameras narrative structure resorts to the
interpretation of both statistically reliable information gained
from the computer-aided process-ing of quantifi able parameters for
each shot and the rich and in-depth descriptive values of
qualitative metadata.
Renowned researchers in the fi eld of statistical stylistic
analysis, such as Barry Salt and Warren Buckland, have repeatedly
and rather successfully demon-strated the immense possibilities of
this method quantifying parameters of style in understanding how
moving image works are constructed.6 Th e ever-growing computing
capabilities and the advent of sophisticated software as Buckland,
quite rightly, points out7 coupled with the closer collaboration of
statisticians and fi lm theorists will, inevitably, lead to a
proliferation of means and techniques and endlessly perpetuate the
analytical process as a whole.8 Th is could, quite possibly, render
it very diffi cult to achieve some kind of consensus among
researchers about
bridge: Cambridge Univ. Press 1987; Graham Roberts, Th e Man
with the Movie Camera, London/New York: I.B. Tauris, 2000; Jean
Mitry, Le Cinma exprimental: Histoire et perspectives, Paris:
Edi-tions Seghers 1974 [Orig. 1971].
5 Catherine Cooke, Russian Avant-Garde: Th eories of Art,
Architecture and the City, London: Acad-emy Editions 1995; Anatole
Kopp, Town and Revolution: Soviet Architecture and City-planning
19171935, London: Th ames and Hudson 1970 [0rig. 1967]; El
Lissitzky, An Architecture for World Revolution, London: Lund
Humphries 1970 [Orig. 1930].
6 Barry Salt, Statistical Style Analysis of Motion Pictures, in:
Film Quarterly, 28, 1/1974, pp. 1322; Id., Film Style and
Technology: History and Analysis, London: Starword 1983; Id.,
Moving into Pic-tures: More on Film History, Style and Analysis,
London: Starword 2006, here pp. 389396; Th omas Elsaesser/Warren
Buckland, Studying Contemporary American Film: A Guide to Movie
Analysis, London/New York: Arnold/Oxford Univ. Press 2002, pp.
102116; Warren Buckland, What Does the Statistical Style Analysis
of Film Involve?, in: Literary and Linguist Computing, 23, 2/2008,
pp. 219230.
7 Elsaesser/Buckland, Studying Contemporary American Film, p.
108.8 Salt reckons the statistical analysis of style could aid in
making aesthetic claims performing
normative functions - by considering in addition to originality
and the degree of fulfi lment of the authors original intentions
the work of fi lm directors that have been infl uenced by a given
artwork. (See Salt, Film Style and Technology, p.37.) Th ese, in
turn, have to be reappraised on the grounds of originality, fulfi
lment of original intentions, and again, the number and importance
of artworks that they have infl uenced.
-
Starvros Alifragkis/Franois Penz54methods, tools and
terminology. Attempting to delineate a particular domain, and
construct an ontology for this specifi c domain, could present a
viable way of deal-ing with what appears to be a problem inherent
to statistical stylistic analysis uniformly applying the same set
of categories to diff erent genres and ceaselessly expanding and
refi ning formal categories.
Ontologies a formal, explicit specifi cation of a shared
conceptualisation9 could function as a means for establishing
communication between a database composed of independent media
items and metadata created to support some specifi c function, or
to describe properties of the resource as used in a particular
context.10 As such, a domain ontology can serve as a mechanism for
unearth-ing what corresponds in the neoformalist approach to fi lm
analysis according to Th ompson to the identifi cation of the
dominant the foregrounding of certain devices that appear to be
all-pervasive.11 Th is research sets off on a shot-by-shot formal
and stylistic analysis of a single movie Man with the Movie Camera
and plans to build from there by considering additional relevant
movies and construct-ing an ontology a fi nite number of relevant
concepts (a formal category/class denoting each one of them, a
description explaining its function and a set of rela-tionships
describing a taxonomy) for a specifi c domain: non-narrative,
non-acted, (primarily) montage-based moving image works, where the
depiction of diverse natural locations of the urban terrain
constitutes an important character12 city symphonies.13 Naturally,
the description of the domain itself is expected to fl uctu-
9 Studer/Benjamins/Fensel, Knowledge Engineering, p. 184.10
Miguel-ngel Sicilia, Metadata, Semantics, and Ontology: Providing
Meaning to Information
Resources, in: International Journal of Metadata, Semantics and
Ontologies, 1, 1/2006, pp. 83-86, here pp. 8384.
11 Kristin Th ompson, Breaking the Glass Armor: Neoformalist
Film Analysis. Princeton/New Jersey: Princeton Univ. Press 1988,
pp. 4344.
12 Character is understood here in the neoformalist sense of the
term as collections of semes, or character traits. (Ibid.,
p.40.)
13 For additional information on the generic features of the
genre, see Erik Barnouw, Documentary: A History of the Non-fi ction
Film. New York: Oxford Univ. Press 1993, pp. 73-81 [Orig.1974];
Colin McArthur, Chinese Boxes and Russian Dolls: Tracking the
Elusive Cinematic City, in: Th e Cinematic City, ed. by David B.
Clarke, London/New York: Routledge 2001, pp. 19-45 [Orig. 1997];
Scott MacDonald, Th e City as Motion Picture: Notes on Some
California City Films, in: Wide Angle A Quarterly Journal of Film
History Th eory & Criticism, 19, 4/1997, pp. 109-130; Carsten
Strathausen, Uncanny Spaces: Th e City in Ruttmann and Vertov, in:
Screening the City, ed. by Mark Shiel/Tony Fitzmaurice, London/New
York: Verso 2003, pp. 15-40; Helmut Weihsmann, Th e City in
Twilight: Charting the Genre of the City Film, 1900-1930, in:
Cinema & Architecture, ed. by Franois Penz/Maureen Th omas,
London: British Film Institute 1997, pp. 8-27; Franois Penz,
Architecture and the Screen from Photography to Synthetic
Imaging:
-
Dziga Vertovs man with the movie camera 55ate as more moving
image projects are considered. Furthermore, this method could be
utilised in analysing other projects by Vertov, because, as Hicks
quite rightly points out, similar stylistic patterns can be
detected in a range of earlier works.14
2.1 Constructing the Ontology: Classes, Subclasses and
Categories
Th e analysis utilised the DVD version of Vertovs movie,
released in 2000, by the British Film Institute.15 Th e DVD has
been sourced from the 1996 restoration of the original BFI copy by
fi lm historian David Shephard.16 For our analysis, the movie was
captured in an uncompressed DV PAL format with Adobe Premiere Pro
2.0. Subsequently, all individual shots were traced, by cutting the
media on the timeline, where the original edits existed. Th is was
neither easy nor straight-forward. For instance, it was diffi cult
to distinguish between shots 437a and 437b medium-long shots of the
cameraman and his movie camera at the foyer of a building, shot
from the interior of an elevator since, in the DVD version, it was
impossible to tell with absolute certainty when the upward movement
of the elevator stopped (shot 437a) and the downward movement began
(shot 437b). Th e mise-en-scne simulates a vertical linear wipe
with top to bottom direction (shot 437a) followed by a similar eff
ect with the opposite direction (shot 437b), while the
mise-en-cadre at the beginning of shot 437a and the end of shot
437b, remains unchanged. In ambiguous cases, such as this one, we
compared our shot list against Petri and Roberts studies.17 Th en,
all 1,694 shots (excluding the credit sequence at the beginning,
the end title and black frames) were numbered and exported as
individual AVI clips (each clip containing a single shot). Th eir
duration varied from one frame to approximately 23 seconds (565
frames). Discrepancies between this and other analytical
approaches, as far as the count of shots is concerned (Petri:
1,682; Crofts & Rose: 1,716; Sauzier: 1,712), can be attributed
to the availability of prints of various lengths. Discrepancies
that involve shot duration have do to with
Capturing and Building Space, Time and Motion, in: Architectures
of illusion: From Motion Pictures to Navigable Interactive
Environments, ed. by Maureen Th omas/Franois Penz, Bristol:
Portland 2003, pp. 135-164. Th e bibliographical research furnishes
the description of the domain with two additional features: a.
dawn-to-dusk narration; and b. episodic narrative structure.
14 Jeremy Hicks, Dziga Vertov: Defi ning Documentary Film,
London: I.B. Tauris 2007, pp 6364.15 ISBN/EAN: 5035673005026.16
http://www.dvdtimes.co.uk/content.php?contentid=5563717 See Petri,
Constructivism in Film, and Roberts, Man with the Movie Camera.
-
Starvros Alifragkis/Franois Penz56the projection speed and, in
instances where source material is on DVD, diff erent methods for
frame interpolation of moving images.18 Finally, one still image
was generated for each shot at a halfway point duration-wise. Th
ese were particularly useful for the annotation of static elements
of form, the segmentation of the plot and the production of tables
that visualise patterns of plot development.
Th e shot-by-shot analysis of Vertovs Man with the Movie Camera
was per-formed in two stages. For the fi rst stage, the analysis
focused mainly on fi lm style (mise-en-cadre, mise-en-scne and
editing), primarily utilising Bordwell and Th ompsons
categorisation. Th is was performed manually, using a table
designed in Adobe Photoshop CS2 9.0.
18 Th e information, such as projection speed and method for
frame interpolation for the BFI DVD version of Man with the Movie
Camera utilised in this research, is not currently available.
Figure 01. Th e manual annotation table accommodates diff erent
aspects of style along the horizontal axis and shots in sequential
order along the vertical axis. Th us, each row of the table
provides coded information dots on a regular grid about a fi nite
number of stylistic characteristics for a single shot, while each
column off ers a way of tracing fl uctuations in the value of a
single stylistic aspect for a sequence of shots.
-
Dziga Vertovs man with the movie camera 57Th e ensuing
lightweight statistical processing and interpretation of the
initial re-
sults, coupled with bibliographical research on Man with the
Movie Camera, provid-ed the basis for mining the relevant key
concepts (classes, subclasses and their cor-responding categories)
for the conceptualisation of the domain.19 For the analysis of Man
with the Movie Camera, aspects of style are pigeonholed in three
major classes: a. mise-en-cadre (cinematography); b. mise-en-scne
(staging); and c. editing.
19 Stephen Crofts/Olivia Rose, An Essay Towards Man with a Movie
Camera, in: Screen, 18, 1/1977, pp. 958; R. Seth Feldman, Evolution
of Style in the Early Work of Dziga Vertov, New York: Arno Press
1977; Id., Dziga Vertov: A Guide to References and Resources,
Boston: G. K. Hall & Co. 1979; Mikhail Kaufman, An Interview
with Mikhail Kaufman, in: October, 11/1979, pp. 5476; Lev Manovich,
Th e Language of New Media, Cambridge, Mass./London: Th e MIT Press
2001; Annette Michelson, Th e Kinetic Icon in the Work of Mourning:
Prolegomena to the Analysis of a Tex-tual System, in: October,
52/1990, pp. 1639; Id. (ed.), Kino-eye: Th e Writings of Dziga
Vertov, Lon-don: Pluto 1984; Petri, Constructivism in Film;
Roberts, Man with the Movie Camera; Yuri Tsivian, Man with a Movie
Camera, Reel One: A Selective Glossary, in: Film Studies: An
International Review, 2/2000, pp. 5176; Id., Lines of Resistance:
Dziga Vertov and the Twenties, Pordenone: Le Giornate del Cinema
Muto 2004; Malcolm Turvey, Can the Camera See? Mimesis in Man with
a Movie Camera, in: October, 89/1999, pp. 2550.
Figure 02. Transcribing the results of the manual annotation of
Vertovs movie. Th e Microsoft Excel spreadsheet, similarly to the
Adobe Photoshop table, accommodates classes, subclasses and
categories along the horizontal axis and shots in sequential order
along the vertical axis. Here, dots have been replaced by a
singular (absolute) numerical value for each category. Th is has
enabled the lightweight statistical processing of the initial
results of the shot-by-shot formalist analysis.
-
58 Starvros Alifragkis/Franois Penz
Figure 03. Th is research singles out three diff erent formats
for visualising the results of the statistical processing of the
accumulated metadata: a. pie charts that describe the contribution
of each category to the total of the movie; b. line charts that
display trends for sets of values over time; and c. tables that
facilitate the detection of recurring patterns. Here, the line
chart combines information for sub-classes: location (green);
number of characters (purple); camera movement (blue); and shot
scale (red) for a sequence of forty shots. Th is research
experimented with diff erent combinations to investi-gate which
categories of style appear to be more relevant.
-
Dziga Vertovs man with the movie camera 59Figure 04. Th e
initial interpretation of the results of the manual analysis
informed the segmentation of Man with the Movie Camera, which in
turn, aided the reconstruction of its formal structure. It could be
said that the movie is roughly about life in a fi ctional Soviet
city from early in the morning till late in the evening. Th e
dawn-to-dusk narration is employed here to provide temporal
continuity to com-pensate for the lack of dramatic, character-based
plot, character development and spatial continuity. It involves the
sequential depiction and arranging of a number of themes or daily
activities (waking up, working, exercising, going to the movie
theatre, etc.) in two themed episodes (a: labour and b:
culture/recreation) marked by the introduction and the epilogue.
Each themed episode contrapuntally combines a number of scenes and
sequences within scenes and concludes with a climactic sequence,
the recapitulation, where motifs that preceded are presented in
abbreviated form. On a diff erent level, the story is told in
roughly three interlocking layers. Th ere is the storyline of life
in the Soviet Union, which is intertwined with the storyline of fi
lm production (shots depicting the cameraman, Mikhail Kaufman, and
his movie camera recording life unawares, and the fi lm editor,
Elizaveta Svilova, reor-ganising the world as a meaningful whole).
Both function as a mise en abyme, when framed within the third
storyline: opening (intro) and closing (outro) sequences of people
watching Man with the Movie Camera at the movie theatre.20 Yet, the
third storyline assumes a unique role; on the one hand, it is an
integral part of fi lm production (thus belonging to the second
storyline), while on the other hand, it concludes the list of
activities that Soviet citizens of that ideal socialist city can
enjoy (thus belonging to the fi rst storyline). Also, the closing
climactic sequence (recapitulation) renders the layers and their
corresponding narrative frames literally indiscernible, thus
merging the three storylines in one story-world. Th e chart
demonstrates the fl uctuation of shot duration (in frames) for the
1,694 shots of the movie. It investigates whether shot duration in
general, and short duration in particular, can act as indicators of
shifts in the formal structure of the movie, marking the end of
sequences, scenes and episodes. Th is understanding of the formal
structure of the movie deviates signifi cantly from both Petris
interpretation who understands sequences showing machines in
operation, sports events, and musical performance executed with
spoons and bottles as integral parts of the same diegetic domain21
and Crofts and Roses analysis which greatly expands upon the waking
and working sequences.22
Th e corresponding subclasses are considered in three groups: a.
qualitative data that cannot be ranked (categorical or nominal
variables); b. qualitative and quantita-tive data that can be
ranked (ordinal variances); and c. quantitative data that can be
ranked, which is approximately normally distributed (scale
variances).23 Th is is because in descriptive statistics, which is
widely used in statistical stylistic fi lm analysis, data of diff
erent nature off er themselves to diff erent kinds of
treatment.
20 Tsivian, Man with a Movie Camera, p. 55.21 Petri,
Constructivism in Film, pp. 7273.22 Crofts/Rose, An Essay Towards
Man with a Movie Camera, pp. 1516.23 On many occasions, one does
not know how to treat variances before one collects the data
and
starts performing some basic tests.
-
Starvros Alifragkis/Franois Penz603. Tools
3.1 Computer-Aided Annotation
Beyond the process of manual annotation, discussed in section
2.1 (Constructing the Ontology: Classes, Subclasses and
Categories), the second phase of the analysis was performed with
Architect Media Tool (BT 2005), an authoring tool for the
production of computer-handled, interactive, digital projects.24 Th
is experimental software was being developed over a period of
several years before it partly formed the basis of a new toolkit
designed for the New Millennium New Media (NM2) collaborative
research project.25 It endeavoured to inspire and realise new media
productions that sought to experiment with non-linear editing for
multimedia da-tabases and the interactive screen. Th is was made
possible by enabling the design of programme templates that
described the overall structure of the moving image work rather
than by producing a single, fi nal cut. In addition, Architect
Media Tool facilitated the generation of multiple programmes
(instances of a programme template) utilising the same set of
resources a single database.
24 Th e project utilised standalone client version v.1.5.22
(released in 2005). Architect Media Tool (pre-viously named Smart
Media Tool and in 2006 Flexible Media Tools) was developed by BT
Ex-act: Future Content Development at Adastral Park, Martlesham
Heath, Ipswich, UK
(http://labs.bt.com/barc/FutureContent.html).
25 NM2 (http://www.ist-nm2.org; and more recently
http://www.shapeshift.tv) is a European Com-munity Integrated
Project, whose principal aims are to produce eff ective digital
tools and experi-ment with novel media genres for interactive
screen and broadband distribution.
-
Dziga Vertovs man with the movie camera 61
Figure 05. Lev Manovichs understanding of the database as a new
symbolic form, a cultural reposi-tory for contemporary society, was
an important point of departure for this part of the research.26 In
a reverse engineering analytical process, Man with the Movie Camera
perhaps the most important example of a database imagination27
provided 1,694 individual media objects (or resources) for the
database: black and white clips with no sound. All individual clips
(each one corresponding to a single shot from the movie) were
imported to Architect Media Tool, where they formed a rich,
non-hierar-chical pool of media.
Th is could happen in three main modes of operation:28 a.
Annotation: Th e Meta-Data Editor, which allows users to annotate
media fi les, was modifi ed to adjust to the aforementioned
ontology; b. Template Production: Th e Template Editor
func-tionality allows users to create programme templates by
constructing individual
26 Lev Manovich, Th e Language of New Media, p. 219.27 Ibid., p.
239.28 Tim Stevens, BT Exact Smart Media Tool: Guidelines for Use.
Martlesham Heath, Ipswich: BT
Exact, 2004, p. 5 [unpublished manual].
-
Starvros Alifragkis/Franois Penz62scenes in sequential order;29
c. Template Population: Reviewing the outcome of a programme
template is enabled through the Content Synthesiser. Th is is where
an instance of the programme template will be assembled and
previewed.
Figure 06. Th e Meta-Data Editor allows the user to annotate
media fi les in two distinct ways. Firstly, with structured
metadata which later on evolved into classes for a domain-specifi c
ontology: a data-model utilised in the retrieval of database items
by the Template Population functionality of the tool. Secondly,
with relationship metadata: a functionality that enables the
sequential ordering of database items, extremely useful for
generating cause-and-eff ect narratives. Double-clicking on the
thumbnail of a media item in the Media Bin window will enable the
Mark-up View for the selected database item. Here, all the
available structured metadata fi elds are displayed. Certain data
can be logged au-tomatically (duration), but most have to be
registered manually, by either typing key phrases or by ticking
boxes for embedded or user-defi ned categories. Architect Media
Tool is an authoring tool, not a statistical package. Accumulated
metadata had to be exported into XML format for further proces-sing
with software, such as Microsoft Excel and SPSS. Th is is also true
for charts, bars and tables.
29 One can change the order of scenes in the Template Editor,
however, during Template Popula-tion, scenes playback in a predefi
ned order: from top to bottom.
-
Dziga Vertovs man with the movie camera 63
Figure 07. Programme templates represent tree-like structures,
whose main building blocks are: a. Combiners (Sequential, Random,
Either/Or and Repeat); b. Filters (Equals, Not Equal, Contains,
Greater-than and Less-than); and c. Funnels. Funnels are used for
setting the maximum number of media items and Combiners describe
the succession of media objects. Th e use of Filters is what
con-nects a programme template to the database items. Filters can
be set to correspond to diff erent values for classes, subclasses
and their corresponding categories. In brief, Filters determine
which database items are summoned, Combiners describe their
ordering and Funnels specify how many of them will play back. Diff
erent combinations of these three structural elements can produce
very complex narra-tive structures, capable of eff ective
storytelling. Th is functionality was utilised to check the
annotation process for inconsistencies and to reproduce Man with
the Movie Cameras formal structure. Th e latter aided the part of
the analysis that focuses on understanding Man with the Movie
Cameras formal structure and linking stylistic patterns to the
structure of the movie.
4. Discussion and Future Developments
Creating programme templates and previewing the results, in
order to test the consistency of the annotation, functioned for
this research as a mechanism for what we have named the process of
disambiguation. Th is did not involve cases
-
Starvros Alifragkis/Franois Penz64where aspects of style were
misidentifi ed (counter-clockwise rotation for clock-wise
rotation). Once traced, those mistakes were corrected without
requiring the implementation of changes that would aff ect the
analytical process as a whole. Dis-ambiguation primarily concerned
instances where subclasses were not developed well enough or
categories were not defi ned clearly enough to distinguish between
events that appeared identical, but were in fact slightly diff
erent. A characteristic example is subclass characters movement
(qualitative, non-ordered variances). Th e corresponding categories
(a. screen left; b. screen right; c. forward [away]; d. back-ward
[towards]; e. up; f. down; g. rotate clockwise; h. rotate
counter-clockwise) are not mutually exclusive (a. a character can
move both screen left and away from the camera; b. there can be
more than one character moving in diff erent directions). Creating
a programme template for subclass characters movement and category
screen left and previewing the assembled clips revealed that, among
the numerous clips that met the predefi ned requirements, there
were a few that in this new con-text, appeared to be out of place
(i.e., clips where a characters screen left movement was countered
by a second characters movement in the opposite direction). Th is
was not simply a matter of correcting the metadata, rather it
required sharpening (refi ning) subclasses and categories so as to
be more sensitive in detecting such phenomena. Th us, the notion of
prominent movement was introduced to aid dis-criminating between
what appeared to be the main movement and, correspond-ingly,
character (although, these two did not always coincide) and less
dominant activities. In addition, issues of prominence are also
considered in relation to issues of association. Th e initial
analysis of Man with the Movie Camera suggests that cer-tain scenes
and sequences are constructed by eliciting analogies in the form
(shape, movement, etc.) between consecutive shots.
Issues of association highlight concerns about the actual
process of logging metadata. Th ese concerns mainly have to do with
whether shots should be anno-tated individually (out of context) or
in sequential order (in context). Th is research experimented with
both approaches: the manual analysis dealt with shots in the order
they appear in the movie, while the computer-aided analysis handled
shots in random order. Evidently, the latter is bound to miss out
on aspects of style, such as time, certain aspects of prominent
movement/shape (by association) and, of course, editing. In
general, annotating shots in sequential order enables more informed
decisions about mise-en-scne. Conversely, annotating shots randomly
privileges subclasses (with their corresponding categories) that
succeed in spotting what one might call inherent characteristics as
contrasted to derived features of a shot and are generally
associated with mise-en-cadre. Th ese metadata appear
-
Dziga Vertovs man with the movie camera 65to be more consistent
and less open to subjective interpretations. Nevertheless, if shots
can be manipulated as easily as Kuleshov suggests in the
descriptions of his infamous series of experiments that led to the
formulation of the same-name ef-fect, any attempt to assign fi xed
meaning to an individual shot and attempt to tran-scribe this
meaning to a metasystem (annotation with metadata) can be seriously
undermined.30 A lot of devices considered typically Vertovian, like
his idiosyncratic usage of the long take (intercutting diff erent
long takes), the emphasis on short duration and the elaborate
structure of sequences (intertwining several activities/actions)
might suggest that modes of progression, rather than individual
shots, are more signifi cant in cueing audiences.
Th e graphical representations of the annotated media and the
segmentation of the plot aided the detection of certain recurring
stylistic and formal patterns. One of the most widely used ones
involves constructing micronarratives sequences within themed
episodes by interweaving two or more activities (driving an
am-bulance and a fi re brigade down the streets of the fi ctional
city). In essence, what Vertov achieves is to fuse diff erent
locations of his composite city.31 Kuleshovs concept of imaginary,
artifi cial landscapes that exist only in celluloid creative
geographies describes the theoretical framework for the fi lmic
reconstruction of Vertovs city: a creative manipulation of the
urban terrain where disparate physical locations construct a
composite cinematic space.32 Boa and Reid suggest that the fi nal
eff ect of the fugal novel as compared with linear novel is of a
spatial network of relationships rather than a temporal succession
of states and events, adding that the more fugal a novel becomes,
the more spatial it will be.33 Th e initial interpre-tation of the
analysis of Man with the Movie Camera suggests that the movie
exem-plifi es the structure of fugal narration, insofar as it
constructs a network of spaces, whose storytelling ability hinges
on the act of interlocking; which could very well constitute a
spatially-arranged narrative. Punctus-contra-Punctum (counterpoint)
montage is introduced here to describe a technique that emphasises
the interweav-ing of two or more activities/locations in a single
montage sequence. Th e interlaced
30 Lev Kuleshov, Kuleshov on Film: Writings by Lev Kuleshov, ed.
by Ronald Levaco, Berkeley/Los Angeles/London: Univ. of California
Press 1974, p. 52; Vsevolod Pudovkin, Pudovkin on Film Tech-nique:
Th ree Essays and an Address by V.I. Pudovkin, trans. & annot.
by Ivor Montagu, London: Vic-tor Gollancz 1929, pp. 164165.
31 Roberts suggests that the movie features fi ve diff erent
locations: Moscow, Kiev, Yalta, Odessa and Donbas (Roberts, Man
with the Movie Camera, p. x).
32 Kuleshov, Kuleshov on Film, pp. 45.33 Elizabeth Boa/J.H.
Reid, Critical Strategies: German Fiction in the Twentieth Century,
London:
Edward Arnold 1972, p.16.
-
Starvros Alifragkis/Franois Penz66activities which correspond to
diff erent locations construct a fi lmic terrain, which, unlike the
continuous spaces of continuity editing, remains coherent despite
its fragmentation. As the analysis suggests, Vertov constructs a
dialectical argu-ment for the ideal socialist city of the future
through the creative manipulation of images of industry,
infrastructure, public space and novel architectural types such as
the workers club.
So far, the interpretation of the initial results of both the
manual and the com-puter-aided analysis of Man with the Movie
Camera suggests that the construction of a domain-specifi c
ontology can provide the necessary tools for the unearthing
relevant to formalist fi lm analytical concepts. Th e proposed
method expands from the statistical stylistic analysis to the study
of recurring stylistic patterns and the reconstruction of the
movies formal structure. Comparing the method, the tools and
results of this analysis against other analytical approaches, such
as the one proposed by Digital Formalism, can determine how eff
ective and relevant the pro-posed concepts in describing and
communicating shared knowledge are within the domain of non-acted,
montage-based moving image works about urban cinematic landscapes
in particular, and the works of Dziga Vertov in general.