This is a brief description of a Web Site and Service I am imagining called (for now) Orgtag. It could also be a feature of a social bookmarks manager.
At Seattle mind camp, I attended a session where Alex Barnett shared his del.icio.us tag profile. I have never been crazy about tag clouds, but I haven’t come up with an alternative. Alex has a very well developed tag cloud (I don’t since I am stubbornly sticking to Lookmarks and haven’t implemented tag clouds). Staring at all those tags, I instinctively wanted to drag them around with a mouse and organize them as if they were those one-word poetry magnets. I also wanted to draw lines between them and label them with relationships like “same as” and “belongs to”.
Another topic being discussed was that of sharing attention data as OPML. This concept is primarily discuseed in relation to RSS feeds, but given the arbitrary relationships I wanted to apply to my tags, and the hierarchical structure of OPML – I have been wondering if OPML might make a good transmission format for personal tag structure as well.
The type of tag web I am envisioning has been experimented on in academia (see WordNet , and maybe Cyc). These type of word relation databases can facilitate text comprehension and inference reasoning. Orgtag would build upon these machine generated relationships with a community voting approach.
Here are the primary features of Orgtag:
1. Import and organize tags to creae a personal taxsonomy
This involves defining relationships between tags.
C# (is a) programming language
Seatle (is geographically inside) Washington
canine (is equal to) dog
The relations could be pre-populated from WordNet, dictionaries, or web crawls.
A single user would start organizing their tags and defining any relationships they see fit. Other users could then piggy-back off of these relationships, chosing which relationships to apply, ignore, or contradict.
2. API that other applications could use to recommend tags based on a content excerpts
I am envisioning drop down menus that allow the tagger to reach a wider variety of tags based on the richness of the relationships defined above. Also, recommendations should based on tag search value as Steve Eisner insightfully points out in his post “Cleaning up the Tag soup” . In the act of publishing, you want to choose tags that will faciliate rediscovery as well as organization, so a high value tag is not necessarily one that has been used before to categorize, but one that has been used before to search.
3. A feedback mechanism
Orgtag imports should not just be one-time imports. Orgtag should accept your RSS feed and continue to import tags and relationships from that RSS feed. This would enable Orgtag to provide highly relevant tag recommendations.