Saturday
Apr232005
Photo Blogs, Wikis, and Memories for Life
Saturday, April 23, 2005 at 7:27PM
After reading The File Manager Is Dead. Long Live the Lifeblog (a brief review of Nokia's LifeBlog application/system), I was moved to try to put down a few thoughts I've had on the topic of annotating photos. (There's also a larger story that I hope to bring out.) First, let's consider on-line tagging systems (AKA Folksonomies if a group is involved), esp. photo-related ones like Flickr. (My favorite non-photo one is del.icio.us/.) I'm not a regular Flickr user, but it seems that it lets you add information to photos, usually in the form of notes and tags. The problem (as I see it) with organizing photos using these two things is that they don't support context, which in this case meas expressing how the photo's subject relates to other things. With photos (and human activities in general), we usually care about WHO, WHAT, WHEN, and WHERE (let's ignore the WHY). But Flickr doesn't explicitly support this. What I propose is a system that merges the following ideas:
For example, pictures of my family on vacation will have:
For photos taken during this vacation, I would use WikiWords instead of Flickr tags for all of the above, with optional narrative. For example, a photo at a party might have these words:
(I assume the WHEN information would come from the image itself, e.g., a digital camera's EXIF information.) It might also have some narration, such as:
Other photos from the same vacation would have similar annotations. Like any wiki, the WikiWords would be made-up at first, but would soon be used to annotation other photos (i.e., emergent classification). At this point you might ask, how are the WikiWords different from keywords? It's a good question, to which I have a partial answer (I could use some help on this). The main ideas (and the ones into which LifeBlog fits) is that a) the words act as links not only to other photos, but to the other non-graphical content of a person's life, and b) the words themselves can have content. In other words, the WikiWords specify the context as a network.
Continuing the example, suppose I also maintained an on-line family journal, with entries in the same format as the photo annotations. Here's a sample of a few (they're in a simple plain text format - separated by dashes, ending with timestamps):
(Note that this would be show in a browser, and each WikiWord would be a hyperlink. Also, you can see there appear to be two types of entries - journal and contact.) My thought is that the network formed by the WikiWords would provide a richer contextualization for all content, including photos. As a result I think this would aid search by providing multiple entry points:
Naturally, getting the richness requires the effort of markup up photos and adding annotations, just like any similar tool. To address this I envision a rapid annotation tool that would let you view photos via thumbnails (ala ACDSee) and easily add WikiWords (or sets of them), perhaps via custom accelerator keys. In addition the annotation tool might support very fast creation of entries (the text between dashes), auto-completion of WikiWords, etc.
Whew! I think that's enough for now, but I will be following this up with more issues, ideas, related work, and how this generalizes to a system for annotating resources of all types, including URLs, books, PIM data, files, etc. For now I'll leave you with a few teaser thoughts/references:
- basic photo management, including uploads and thumbnail views
- semi-structured annotations using WikiWords interspersed amongst plain text
For example, pictures of my family on vacation will have:
- WHO - zero or more people taking the trip (family), and zero or more people at the destination (e.g., friends)
- WHAT - the vacation event itself, and possibly a special event (e.g., dinner, party)
- WHEN - the date and time
- WHERE - scenery from the vacation destination
For photos taken during this vacation, I would use WikiWords instead of Flickr tags for all of the above, with optional narrative. For example, a photo at a party might have these words:
BobSmith
SueSmith
AlanJones
Florida2005Vacation
CocoaBeachFlorida
JonesHousehold
(I assume the WHEN information would come from the image itself, e.g., a digital camera's EXIF information.) It might also have some narration, such as:
Here we are at the Jones's house during the great party they had in our honor. You can see Sue hamming it up on stilts.
Other photos from the same vacation would have similar annotations. Like any wiki, the WikiWords would be made-up at first, but would soon be used to annotation other photos (i.e., emergent classification). At this point you might ask, how are the WikiWords different from keywords? It's a good question, to which I have a partial answer (I could use some help on this). The main ideas (and the ones into which LifeBlog fits) is that a) the words act as links not only to other photos, but to the other non-graphical content of a person's life, and b) the words themselves can have content. In other words, the WikiWords specify the context as a network.
Continuing the example, suppose I also maintained an on-line family journal, with entries in the same format as the photo annotations. Here's a sample of a few (they're in a simple plain text format - separated by dashes, ending with timestamps):
----
started Florida2005Vacation! SueSmith got the flu
(2005-04-01 14:05:40)
----
party at JonesHousehold (see the pics). fun! talked about SavingTheWorld with AlanJones
(2005-04-02 13:15:30)
----
AlanJones
Alan X. Jones
23 S. North Street
Cocoa Beach, Florida, 32323
321-321-1234
alan@gmail.com
married-to: JanJones
(2004-03-23 03:23:43)
----
...
----
(Note that this would be show in a browser, and each WikiWord would be a hyperlink. Also, you can see there appear to be two types of entries - journal and contact.) My thought is that the network formed by the WikiWords would provide a richer contextualization for all content, including photos. As a result I think this would aid search by providing multiple entry points:
- WHO - content search using WikiWords
- WHAT - ""
- WHEN - date search
- WHERE - content search using WikiWords, and geographical search using map, etc.
- graphical search for related items using link perusal (think Touchgraph for your photos, with thumbnails!)
Naturally, getting the richness requires the effort of markup up photos and adding annotations, just like any similar tool. To address this I envision a rapid annotation tool that would let you view photos via thumbnails (ala ACDSee) and easily add WikiWords (or sets of them), perhaps via custom accelerator keys. In addition the annotation tool might support very fast creation of entries (the text between dashes), auto-completion of WikiWords, etc.
Whew! I think that's enough for now, but I will be following this up with more issues, ideas, related work, and how this generalizes to a system for annotating resources of all types, including URLs, books, PIM data, files, etc. For now I'll leave you with a few teaser thoughts/references:
- I'm trying this approach out at work as an experiment for my professional log, using (prepare yourself) emacs. No support for photos yet.
- structuredblogging.org and PubSub's structured blogging initiative
- Collaborative knowledge gardening, Comments: Flickr Graph, Tagging, On Personal Information Management
- Organizing Thoughts into Sequences, Hierarchies, and Networks
- a cool photo site
- Using Properties for Uniform Interaction in the Presto Document System
- Issue: Topic hierarchies? Example: JonesHousehold is in TownOfCocoaBeach in StateOfFlorida in CountryUnitedStates.
- Issue: word/tag sets -> topics. Classifier to help recognize/name these?
- WikiWeblogPIM and Jot
- MyLifeBits Project
- Grand Challenges in Computing (British Computer Society), esp. the section on Memories for life: managing information over a human lifetime
Reader Comments (4)
That's an excellent idea. Good thing about that is one could link up pieces of the puzzle that ended up being in the final project. A montage of 5 images wouldn't need all 5 images recopied into that project directory (like it is done now at work so there are many duplicates across the whole server), but linked instead.
I wonder if one of the things one could search on is by use. Was the photo part of a project? Or are there other versions of it?
I have a problem at work finding versions of one image a lot. At home it's not so much a problem ... yet.
You're right on, Jessica. And search by use is exactly the kind of advantage this approach would allow. Really, it's how we think of things as humans, which inspired the whole hypertext idea in the first place. Utilizing the relationships between information items (including pictures) is a kind of link analysis, AKA social network analysis, and it allows all sorts of cool search and navigation.
So how should we go about building this? :-)
matt
I've discovered Jot about two weeks ago, and I'm looking forward to use it as a wiki + software test management tool.
About the digital data deluge we're living, I do agree with you: software has yet to evolve in order to help us (users) to navigate through all the data. So, tags/keywords/WikiWords could help. That's a great concept, now I'm wondering about possible implementations.
I'd love to hear how Jot works for you, Ricardo. We're thinking of adopting it at work for a project. Thanks for the comment.