[Nettime-bold] Onewordmovie: an internet-based installation

[Felix Stalder <felix@openflows.org>]

One Word Movie
An Internet-based project by Beat Brogle

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+ March 06, 20:00 at [plug.in]: «onewordmovie» Opening: Presentation of 
+    the Project and Prototype +Installation by the artist.
+ March, 7 - 23 «onewordmovie» at the [plug.in] Wed - Sun 14-18 Uhr
+ Starting 7. March «onewordmovie» Onlineversion accessible under:
+    www.onewordmovie.ch. Preview-Version for the Media 27. February:
+    www.xcult.ch/onewordmovie 
+ www.youplugin.org / office@iplugin.org
+ St. Alban-Rheinweg 64
+ CH 4052 Basel / Switzerland
+ Tel. ++41 (0) 61 283 60 50 / Fax. 51
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++



About the Project
============================================================================


One Word Movie
	From a world of images to an image of the world.


«One Word Movie» is an on-line platform which organizes, based on 
user-supplied terms, the flood of images on the Internet into an animated 
film. A word turns into images, images turn into a movie.

Using a specially programmed search engine, users can call up images from the 
Internet which match their search term. The project's search engine is built 
on top of the most popular image search facilities available on the Internet. 
Supplied with a search term, the engine produces first a «hit list». This 
list can be several thousand images long, depending on the term. The images 
on this «hit list» provide the «raw material» for the movie. Following the 
ranking on the «hit list», the images are animated into a film in real-time, 
following a fixed and predetermined score, which consists of series of 
interwoven loops.

Each film has an individual trailer, displaying the search term as the title 
and each film lasts until the «raw material» is used up.

This project plays with the tension between on-line and cinematic approaches 
to images. Consequently, it functions as well online as in real space. Hence, 
it will be presented in two versions: as browser-based application and as an 
installation. The on-line version is accessible via a normal internet 
connection and standard browser at http://www.onewordmovie.ch. The 
installation, closer to the cinematic experience, consists of three video 
projections onto three screens.

On-line version
Like with any search engine, the user enters a search term into a web form. 
The engine then collects images from its database and creates a hit list. 
This happens in the background, invisible to the user. Based on this hit 
list, a film is being created and streamed in real-time over the Internet to 
the user. The server can handle several streams, hence several users, in 
parallel. Each film will be generated instantaneously using the latest images 
from the web. 

Installation
The central spatial element of the installation is a triptych consisting of 
three screens, each showing a different data projection. Each projection is 
one animated film, hence the user can see up to three films in parallel. The 
installation creates a cinematic situation, i.e. the projections dominate an 
otherwise darkened room. In this room, in addition to the screens, there is 
also a control monitor which indicates which search term each movie is based 
on. Even though the situation is cinematic, the audience is not passive. 
Rather, the control monitor also allows users to enter their own search 
terms, hence directly influence, though not fully control, the films that are 
shown.

The installation consists of three projections so that the users can relate 
the different streams to one another and thus explore yet another level of 
Internet-specific image-word relationships. One can imagine, for example, 
that the same search term entered in different languages creates different 
films, or that two words whose meaning is very similar are associated with 
very different image clusters.

The best context for the presentation of the installation are media 
festivals, galleries or museums. These institutions still struggle with the 
question how to integrate the Internet into their traditional exhibition 
spaces without recreating the individualizing situation - one machine = one 
user - that we all know from the office or home use of the computer. The 
installation version of the project can help to overcome this difficulty by 
choosing a format that is adapted to the exhibition spaces as much as it is 
to the Internet.


Discussion
============================================================================

The set-up - create a film according to predetermined score - has a certain 
kinship to the structural film of the 1960s and 1970s. There, as in this 
case, the role of the individual author is minimized, without ever 
disappearing completely. It shifts from the selection of images and scenes to 
the creation of an algorithm that guides the selection. In «One Word Movie» 
the remaining fragments of authorship are split again, into those pertaining 
to the author of the set-up and those pertaining to the user who initiates a 
particular search.

However, such similarities to conventional (experimental) film are outweighed 
by differences and these differences are the actual core of the project. They 
stem from the altered status that images have online and the strong influence 
of the specifics of the technologies employed.
The strong and influential particulars of the technology relate to the fact 
that a semantic search strategy is employed to solve what is - from a human 
perspective - a pictorial search problem. Instead of analyzing images 
according to the visual content, as humans do, the machine is searching for 
semantic terms which appear either in the image's file name or in the text 
closely associated with it. This approach to image searching is the most 
widely used today, including the large commercial image search engines. 

All attempts to analyze visual content automatically have failed so far. One 
program to look for pornographic images, for example, the «naked people 
finder», was searching for colors resembling that of naked (white) people and 
analyzed the shapes of these colors. Two pink cylinders, say, were 
interpreted as two naked extremities. The search precision, however, was too 
low to be of any use. Particularly often, images of desserts were 
misidentified as pornography. The  project is no longer continued.1
Because of such non-trivial challenges to automatic content analysis of 
images, it is highly likely that the semantic approach currently used will 
remain predominant for quite some time.  No matter how adequate it may be, it 
is the way most people will search for images. It makes little sense to 
critique it, since there is no alternative available. One Word Movie explores 
its ramifications. Two questions come to the fore: What is the relationship 
between the file name and the image content and, secondly, what kind of 
selection is being made, supposedly representing our search interest?

The question that has figured prominently in modern art - the relationship 
between image and text - is posed anew, this time a practical and urgent one, 
directly affecting the habits of millions of Internet users.

The fast-paced sequences of the images appear to the viewer initially as 
random. The human cognition, focusing on image content, has a hard time 
understanding the machine's selection criteria. There's a fundamental 
disconnect. As little, or as well, as we cannot infer the file name from the 
image content, the machine cannot infer the image content from the machine.
However, these flickering images contain rich cultural information. As an 
example, the film visualizing the search term «me» contains patterns that are 
revealing the current state of (online) culture. The majority of images we 
see are of young white males, females appear only after a while and 
non-whites even later. Intermixed with this quite expected standard imagery 
is a surprising number of x-rays and CT scans offering new, highly 
technological views on the individual body. A lot of people apparently take 
these images as good representations of themselves, which we can assume since 
they call the file "me" or associate the term closely to the image. This 
implies the question how far new visualization technologies are changing the 
view on the human body. The film suggests, this is not only a problem 
affecting some specialists, but changing perceptions of the (online) 
population at large. What is particularly striking is that there seems to be 
such trust that these complex, heavily mediated images are indeed good 
representations of an individual person.

The flood of images that «One Word Movie» channels into a movie are at the 
same time without meaning - they haven't been assembled by a cognitive 
subject - and full of unintended meaning if viewed by a perceptive person. 
The lack of meaning is heightened by the difficulties of a semantic image 
search in a non-semantically classified image pool. Despite certain 
similarities, the Internet does in this regard not resemble a structured 
library or an archive.

The meaning of the images lies not in the individual image - which, after 
all, passes by so quickly that it can barely be detected - but the patterns 
that this high-speed view of image groups reveal. This is a cognitive mode 
quite typical for the electronic culture. McLuhan understood this clearly 
when we wrote:

Information pours upon us, instantaneously and continuously. As soon as 
information is acquired, it is very rapidly replaced by still newer 
information. Our electrically-configured world has forced us to move from the 
habit of data classification to the mode of pattern recognition.2

«One Word Movie» forces us, who are used to look for narration, to shift our 
perception. In these films, there is nothing narrative, even if we tend to 
detect some fragments of a narration from time to time. But any narration 
rests on a narrative intention, which we cannot assume a machine to have. 
However, if we are able to free ourselves from this search for narration, we 
can explore hidden informational patterns. The flickering films open our view 
on to a hidden «psychogeography» of the Internet.


Technology
============================================================================
Input of search terms and search parameters and output of the animated movies 
are delivered in standard web browsers - like e.g. Netscape or Internet 
Explorer - on the computer of the viewer of One Word Movie. The One Word 
Movie application has active components on three levels: a shockwave 
application in the browser of the viewer, a script on the web server and the 
image search engine of Google.

After the web page with the embedded shockwave application is loaded, an 
input field for the search term is displayed, along with different options 
like play-speed, image repetition, image size and image-loop length. By 
clicking on the search button, the search term is sent to a script on the web 
server which transmits the term to the Google image search. The results page 
from Google is parsed and interpreted by the script to extract the image 
URL's which then are sent back to the shockwave application.

The application is loading the images from the internet to the computer of 
the viewer and standardizes the images since they exist in different formats 
and sizes on the web. On the interface of the application the growing 
animation can be viewed. In the background, invisible to the user, new images 
from the Internet and new image URLs from Google are loaded and inserted 
constantly into the animation.

The image search itself and the transfer of the images from the Internet take 
place with a certain delay, wherefore the beginning movie is only a few 
images long but grows up to the full length within seconds.



_______________________________________________
Nettime-bold mailing list
Nettime-bold@nettime.org
http://amsterdam.nettime.org/cgi-bin/mailman/listinfo/nettime-bold