Kevin Cheng  

Search is Not All There

October 15th, 2004 by Kevin Cheng :: see related comic

With a theoretical limit on the amount of information we can possibly consume, finding the information both on the internet and on your own computers can be a daunting task. Recently, there’s been a lot of attention on search. Of course, there was the Google IPO, Amazon’s A9, Ask Jeeve’s redesign etc.

Beyond the online searching, a crop of desktop tools have been released. Mac users have LaunchBar and QuickSilver while Windows users have their answer with AppRocket. Microsoft purchased the uber fast search plug-in for Outlook, aptly dubbed Lookout. In Wired, Steve Jobs claimed that searching rather than sorting is the way of the future when unveiling their Spotlight search tool. Not to be outdone in anything that involves the word “Search”, Google now has a beta version of desktop search which scours most Microsoft Office files, text files, web cache, e-mail and AIM chat transcripts.

With all this fuss, one would think search was the be all and end all. Sadly, even with all these options on the market, I’m left feeling that there’s still a lot more work to be done. Why?

Mega Metadata

One major limitation of all of these applications is that they require a huge need for annotation. Spotlight, Jobs’s search software mentioned in the aforementioned article, searches through the text of documents but also parses through a slew of metadata like file type, creation date, file size, etc. Therein lies the key: for anything other than text files to be useful, we need to annotate the files with metadata.

For some, tagging every single MP3 with ID3 data is a weekend past time. Imagine doing this for every graphic, photograph, video and audio file you own. Only then could a search be comprehensive enough to become the primary tool for navigating files in your computer.

The Need to Browse

To continue with the music file analogy, I don’t always know what music I want to listen to. I may have some guidelines like “mellow” or “something like Cafe del Mar albums” but my task is to browse. In such cases, search can never replace sort.

Solutions at Hand

Fortunately, many tools are cropping up to fix the annotation problems of both images and music. For images, we have the increasingly popular Flickr community and the brilliant ESP Game which pits two players against each other to find common keywords for a set of images. For music, community driven programs like MoodLogic have completely changed the way music is handled. With the community’s music profiles, I rarely need to actually entry music data and can find music that “feels like” a song or genre or even just specify a mood.

All of these solutions have one thing in common: community. To require everyone to tag their own files is unwieldy and near impossible so having a community to do so makes sense. The major hurdle, however, is privacy. Having a community tag music is one thing but images and videos are usually more personal. Those who use Flickr are obviously less apprehensive about sharing their photos but personally, I don’t want to share ALL of my photo catalogue with the rest of the world. I’m not sure there is an easy solution to this problem - “tagging” comes to mind as an approach but still requires a lot of work, especially for the amount of pictures people take nowadays. Unfortunately, it’s also the most pressing problem. My most frequent task of late has been to find a certain photo to show someone and my solution has been to have a clear album structure. Search is a long way off from solving this issue to the point that I can remove that hierarchy.

Other approaches to search include the use of hybrid models. Endeca, for example, utilizes search and browse simultaneously. Some are experimenting with browse/filter models which can be effective depending on the task.

Not All There

Search seems to be the buzz at the moment as the latest and greatest form of interaction. If it is, I think it has a ways to go before it can be hailed as such. Stay tuned for Tom’s discussions on whether search is a better paradigm in the first place.

9 Responses to “Search is Not All There”
Anonymous wrote:

Re metadata: Yes but it does allow us Liberians to produce many a specialist thesauri

Josh-H wrote:

While manual metadata tagging is labor intensive, automated tagging could help reduce the workload. One system for tagging everything would break down fast, do not want to try and build a framework that relates pictures of my grandmother to academic papers. However automated tagging for specific types of related data objects could alleviate this problem while providing the benifit. Rather than tag everything to try and relate it all, wouldn’t domain/task/type specific tagging be a better place to start? Would also be easier to write programs to auto-generate tags, saving manual work. Thoughts?

Eric Svoboda wrote:

I’ve been thinking about a system that would attach information to digital photos. It would work something like this: the camera is GPS-aware. Locales (tourist destinations mostly, but possibly private homes, schools, etc.) would provide up-to-the-minute metadata (any special events occurring at the moment, temperature, weather, etc.) to a central database. When the camera takes a picture, it makes a note of it’s geographic coordinates and universal time (via GPS). The camera (or some intermediary technology) can query the database and attach the relevant metadata to the image.

I imagine that looking out maybe 5-10 years, that computers will automatically collect metadata by analyzing the context and environment in which the information is created. I guess this already happens to a small degree. Perhaps your computer will also observe your conversations, taking note of things like emotion and gestures. “Computer, remember a few years ago - around Halloween - I think I was with Bob near the art museum - I think I took a picture of his dog catching a frisbee. Could you find that for me?”

Nico Macdonald wrote:

This post would appear to relate to the Perfect Search Engine Interface post, which discussed visualisation, among other themes.

Rebecca S wrote:

See also various projects from UC Berkeley’s SIMS - Zooke, a camera-phone-based photo annotation game; various other efforts out of “garage cinema” on automated metadata tagging of media, and Flamenco, an image searching/browsing interface.

Kevin Cheng wrote:

Peter Merholz has a post about Metatagging for the Masses on Adaptive Path and a discussion about it on his blog.

Moi wrote:

> Perhaps your computer will also observe your conversations, taking note of things like emotion and gestures

Christ! I hope it doesn’t remember the gestures I make at my computer, it would get very angry.

Kevin Cheng wrote:

Wired has an article on meta tagging photos: Point, Shoot, Kiss It Goodbye

cheap bass guitar wrote:

Interesting. I am thinking about getting a cheap bass guitar from this site. Do you think the prices are ok?

Leave a Reply

OK/Cancel is a comic strip collaboration co-written and co-illustrated by Kevin Cheng and Tom Chi. Our subject matter focuses on interfaces, good and bad and the people behind the industry of building interfaces - usability specialists, interaction designers, human-computer interaction (HCI) experts, industrial designers, etc. (Who Links Here) ?