Q: Why archive Google image searches?
A: The answer to this question is crucial and is at the very crux of why we made the Tracer. While its potential use will be only fully understood once the initial traces have been executed over an extended period of time, we have a few ideas of what might be extrapolated from the gathered data. As a result of doing manual traces, meaning taking screenshots of searches over several months, we saw how images move in rank in relation to media events. It was interesting to note what images became iconic or representative of a certain event, and what images did not. We started to wonder about those images that were lost or taken offline. In archiving them, they could perhaps reveal something about the public perception of certain event. Next to this, there are also observations to be made about what happens through different language searches and how that represents local, cultural and social differences.
Q: How did this project start?
A: It began September 2005 when De Geuzen held a three-day lab called Historiography Tracer in Oostende, Belgium. The lab brought together people who were interested in the nature of online media images. Attendees gave presentations addressing a variety of issues such as, the difference between online media images and traditional analog photographs, the distribution of images through the web and the impact search engines have on rendering images visible or invisible through their filtering. Some talks also touched upon concerns like the rapid flow of visual information due to online immediacy and its influence on future understandings of history.
Q: What will you be working on for the next release?
A: For a large part this depends on the results we will be seeing in the coming months. We are looking for ways to map patterns of change, meaning we are looking at ways to identify what images are especially “hot”, and at which days or in what periods we see lots of changes. Of course we are most of all interested in how specific search terms behave, and how changes in image results relate to events taking place elsewhere. We are working on the automation of the whole system from the downloading images to the publishing of html pages; at the moment each update requires manual work. We are considering whether it would be interesting to allow visitors to trace words using a web interface; also we would like to add a page of detailed information per image.
Q: How can I use the Image Tracer myself?
A: The tracer script will be made available for download soon after we have gone through this initial test phase. The script is operated from the command line or shell, and archives image search data on your computer; it also downloads and resizes images for you. Currently it does not automatically generate html pages. We will be working on a “how-to guide” to make distributed use possible in the near future.
Q: I saw three identical images but they all mention: ‘found 1 time’. Shouldn’t this have been ‘found 3 times’?
A: If you look at the image data, you will notice that the image url and/or image name is different for each of the images. Each image on the web is identified by a unique url; in other words, no more than one image can be parked at any single address. This means that even when the exact same .jpg is placed in two different folders, a search engine considers them to be two and not the same. Consequently, when an image file would be overwritten by another file with the same name (for example image.jpg was first a picture of Ariel Sharon, but was later replaced by another file, also named image.jpg but this time it is depicting Osama Bin Laden), there is no way the Image Tracer can detect anything has changed. Here is where you see search engines are calculating but not thinking in the sense that a human could easily see the difference.
Q: How is the Image Tracer different from the Wayback Machine at archive.org?
A: The main difference is, that the Wayback Machine archives web pages, while the Tracer archives image searches and the images themselves. Furthermore, the interface of the Tracer allows you to layer results on one page and see changes over time in a glance. And once the Tracer can be downloaded, archiving will be distributed and not centralized.
Q: If I search for Radovan Karadzic (or any other term on the list) in Google, the search results differ from the ones presented in the Image Tracer.
A: If you compare the Tracer results at ../tracer/Radovan_Karadzic to the ones in images.google.com/images?q=Radovan%20Karadzic, the results could be different for two reasons: either there has been an elapse of time between your search, and our last uploaded search results, and in the mean time changes have happened. But we did notice that the search results found directly in the Google database, were slightly different from the ones presented to us through a webbrowser. We are not sure yet why, and are still investigating…
Q: Does it matter if I have set my Google preference set to ‘safe search’ or specified language preferences and things like that?
A: No, the Image Tracer website is a collection of static html pages and the data represented in it is not influenced by any of your local settings. If you are running the tracer script on your local harddisk yourself, your cookies, preferences etc. would still not influence the results because the Image Tracer script accesses Google’s database directly.
Q: How does the Image Tracer differ from Google’s News Archive?
A: While the Tracer can track what is topical, it is not a news source. It is tracing the way images appear on the web through Google’s filter.
Q: How is the Image Tracer different from Google Image Search?
A: The Tracer software uses information gathered by Google Image Search, but saves the results and presents them in a specific way. It makes it possible to look at changes over longer periods of time, and also creates a distributed archive of images that makes analysis and reflection possible.
Q: Can you look for other words as well please?
A: We could, but right now we are interested in this specific set of words. Eventually, you will be able to download the script to your hard disk. Our aim is to make the Tracer a useable tool for others to research their own areas of interest.
Q: Why did you decide to search this specific set of terms during the test phase?
A: We selected a spectrum of words ranging from key media contenders, such as Bush and Osama Bin Laden, to particular nations making the news like Iraq, Iran and Israel, and more general terms in circulation like terrorism, riots and refugees. While some words are specific to particular conflicts , other terms move across political situations. In a sense, based on what is happening now, we had to do a little news predicting and wanted to create a kind of cosmology of intersecting terminology. Getting slightly more detailed, names like Blair and Radovan Karadzic have been put in because it is likely that Blair will step down and Karadzic will eventually be caught so at one point their images will experience a shake-up. Both America and US are in the list because we noticed a great difference in what those two words rendered despite the fact that they refer to the same place.
Q: What information do you archive about each image?
A: It’s url, it’s status and its rank. This information is stored in a database together with the date it was gathered. Information that may not seem meaningful at a given moment, might become interesting through repeated searches.
Q: I am looking at the Image Tracer in Safari / Internet Explorer and I do not see any image data…
A: Certain browsers unfortunately interpret parts of the Image Tracer style sheet incorrectly. You can use Firefox instead, or have a bit of patience until we found a way around!
Q: How long does it take to do a tracing?
A: The first time the tracer script is run on a term, it will take roughly an hour. Each term added, adds another hour. This is because all images need to be downloaded to a local computer before they can be re-sized. Subsequent traces will take less time because there will be only few new images found. That is unless, a large media event occurs which categorically disrupts the image of a word. For example, five years ago, “Twin Towers” would have rendered very different images than it does now. This is what we would consider a radical and sudden image shift.
Q: Why does the same search term bring up such different images?
A: If you think of the world wide web as a multi-authored archive with each archivist having his or her own labeling system, then that can sometimes render some very disparate results. For example, if you query the word “violence” you might pull anything from Amnesty International, to a local heavy metal band’s fan site, to a pornographic fetish url. Furthermore, the awkward way images are identified through contextual data, sometimes produces freak effects.
Q: What about image copyright?
A: Google claims no rights over the images it searches and nor do we. Where Google provides image searches as a commercial service, our archiving tool is meant for research purposes only.
Q: What possible changes can I expect to see with each tracing?
A: Images may shift in rank, new images may appear and some may appear as broken or missing.
Q: How does an image search engine work?
A: A search engine cannot look at pictures, so it looks for contextual data instead. This includes image name, image folder or alt tag plus any other text surrounding the image. An automatic script (‘crawler’) virtually visit each web page available on the World Wide Web and gathers information; it means that only images that are included in webpages, are indexed [?]. All relevant data about the image, plus a tiny version of it (thumbnail), is saved in the search engines’ database, so that it can be quickly accessed when you type in a search term. Depending on the search engine, each page of data is weighed and in that way the rank (and therefore retrievability) of images is defined.
Q: Can a search term be submitted in another language?
A: Yes, and that of course renders very different results and might give insight into cultural/linguistic differences.
Q: Why did you decide to use only Google Image Search and not other search engines?
A: At the moment there are only two image search engines still active: PicSearch and Google Image Search. Alternative search engines may use their own software to search through texts, but when it comes to finding images they rely upon data from these two sources. Because Google is increasingly dominating our experience of the World Wide Web, we find it necessary to look closely at how files are lost, found and ranked under their filtering influence.
Q: Can you trace any other words for me?
A: No, but you can run the tracer yourself. Go here and follow the instructions.
If you have comments, corrections or additions to the answers given in this F.A.Q., please contact us at info @ geuzen.org