When you look at a Learning Tree like Athena. You will notice something is different. The branches look strange, because each link is a tag.
Also, they appear to be in random order. The tags are ordered by weight, structurally, not semantically, by how many posts they are associated with. Some of you may also notice that tags appear in multiple places, but often with a different number of posts. OK, we agree with you, that’s kinda strange, but there is a reason…
Building trees from tags is a great idea, with some gnarly problems. One is, tags are flat, not hierarchical, which is why they’re tags & not categories!
. Yet, we want to build categories from tags. When we hand-create categories, we try to be logical & arrange our terms in a logical order. When we build trees from tags, we are not there to tell our software how to organize those tags, so we devised a rule…
Each tag can only join the branch if it has posts in common with the tag before it. This establishes a context chain between branch tags, what we call Progressive, reticulated or articulated context chains. We start building a branch by ordering tags by the full number of posts they’re linked to, then, see if any post a tag is associated with is also associated with its previous neighbor.
So, the sequence of tags in a branch is complying with a logical, but not obvious rule. This isn’t perfect, but it’s a start. We intend to build more powerful trees in the future.
You need to look at these trees in a new way, as a tool for discovery, not search. Each branch is a Pathway, a slice of content, not a search guide. In a traditional e-commerce site, you drill down to Women>Hiking>Boots & Shoes>Shoes to browse day-hiking shoes. In a Simili Learning tree, you may see California>San Francisco Bay Area>Rancho San Antonio, Pacific Northwest>Pacific Northwest>ANgelina Trailrunner. They are names, not nouns, and clicking the leaf links may give you posts about trail shoes, or about hiking in the Bay Area. This is better suited for someone considering taking up day hiking who may be new to the Bay Area than someone looking to buy a pair of shoes. It is not geared for commerce, but learning what yuo need to know to go day-hiking in Northern California. This is the price we pay for creating automated categories, we lose efficiency but gain knowledge.
How is a Learnin Tree useful then if a traditional E-commerce Category tree will sell more shoes? A traditional category tree serves Website owners by bringing in revenue & readers by letting them buy shoes, but nothing else. There are many sites where you can buy shoes, but the traditional method is designed to get you to spend money on that site alone. It’s competitive. Our approach breaks this rigid model by mixing content, even across websites. This exposes other aspects of a website, or websites that otherwise would remain hidden by the impetus to draw readers in & get their money. We can write many articles about this, but the most compelling point is, we trade action for engagement, we trade a captive audience for whitelisting. When readers can whitelist multiple websites, we have the basis for building a knowledgebase, for the more websites we have covering particular concepts, the more corroboration we get, the greters the depth & width of coverage & the more novel the tangents to explore. This makes the Website less siloed & more a source of knowledge. It also establishes relatedness between content spanning many websites over the net, so is the basis for an internet-wide discovery space, MemeWeb and the Internet Library for Humanity, Terra Cognos
That said, sometimes you just need to browse women day hiking shoes. If you need to search for a particular post, use the Classic tree.
Leave a Reply