Affinitomics™ performs two basic functions regarding tags: 1) it breaks-down tags into three types, and 2) it allows them to be attenuated with a value or amplitude.
These types essentially existed before, but they were “generic” and not as useful. In A.I. circles, a flat tag structure is called a “bag-of-features” or “bag-of-words.” That means that tagged information shares the same value. For example, a post tagged with “dogs” has the same value as every other post tagged with “dogs.” Some systems try to ascertain similarity based on a count of how many tags are the same. Such systems are extremely prone to “noisy” answers – like internet searches which return irrelevant data.
Affinitomics™, on the other hand, breaks tags down into three types; 1) Descriptors, 2) Draws, and 3) Distances.
Most tags are descriptors. They describe a feature of the information they represent. Using our example “dogs,” an article about dogs would likely be tagged as “article, pets, dogs.”
Some tags are draws. This means they infer a positive relationship, thus attracting what “goes with” or “likes” the post. In the case of the dog article, draws might be “+dogs, +fetch.”
A few tags are distances. These infer a negative relationship, thus repelling “features and elements” that the post (represented by the tags) “doesn’t like” or “doesn’t go with.” In the case of the dog article, a viable distance might include “-cats.”
Put together in an Affinitomic™ syntax, the post “dog” would look like this: Dog Article} article, pets, dogs] +dogs, +fetch, -cats.
Furthermore, both draws and distances can be given an amplitude (value) from -5 to +5. Draws are given a positive amplitude, while Distances are given a negative amplitude. This tells Affinitomics™ how great or severe the like or dislike is. For example, +dogs5 is a 5x higher attraction to other dog articles, and -cats5 is a more severe dislike than -cats.
This means that an article tagged with Affinitomic™ elements can be evaluated for context, value of content, and relevance to other articles in the collection.
With nothing more than a simple instruction, articles (post, pages, custom post types) become self organizing within the collection based on 1) draws to other articles, 2) similarity to other articles, 3) distances from other articles, or 4) context of the articles.
It’s interesting to note that distances are not stop-words, or exclusions. In fact, an article tagged with -cats5 will have a strong positive relationship with another article tagged with -cats5. In that case two negatives make a positive. Or put another way, “The enemy of my enemy is my friend.”
Essentially, Affinitomics™ gives WordPress the tools to enact an A.I. construct similar to that of Support Vector Machines (SVM) – currently one of the most widely used forms of A.I.