Basic Vocabulary Algorithm

Honestly, this is the simplest part.

Start with 10K of the most commonly used nouns and verbs (AKA "proposed terms"). Generate an overall 'score' (in terms of weighted "degrees of separation") from each Type (and the Type's other 'true meanings') to each proposed term. If a Type scores in the top 50th percentile for a given proposed term, that term contains the element (or the bit) that the Type represents.

When each proposed term has been 'vetted' by all eight Types, its basic index is completely determined. This process is "exhaustive but straightforward", which makes it an ideal task for a computer.

Definitions of terms used above:

The 'weighting' algorithm would run as follows:

  1. Each proposed term would be checked for the first "Direct connection" (Type in same sentence) with the applicable Type (Self, for example)
    1. for efficiency sake, same sentence, paragraph and page might be checked simultaneously, but only same sentence would matter at this point
  2. The largest group of terms (found or not found) would be checked for the first connection with the next 'true meaning' of that same Type (probably Supremacy)
    1. "trueness of meaning" is still a judgment call
    2. a different order of 'true meaning' checks might provide better term grouping
  3. The sub-group that contains the fiftieth percentile would be checked with the next true meaning (Ownership perhaps?)
    1. each iteration checks a smaller number of terms
  4. Repeat step three until the 'need' and 'don't need' groups are of equal size
    1. this may not require checking all direct connections or any indirect connections
  5. When done with one Type, Repeat process with next Type
reiterative cuts

The only requirement is a repository of connected words (like the internet or a huge, general purpose Wiki) and the computer resources required to 'sift' it. If anyone knows of any organization that would be willing to support this project, let me know. Unfortunately my resources are limited. (I don't have the space to *download* Wikipedia, much less process it.)

I'm only specifying ten thousand words, because "Non-descriptive conversational phrases" should fill most of the remaining six thousand basic slots. Common conversational phrases (and other handy terms) should be supplied by users to fill their own most usual or pressing needs. Personal vocabularies should be used for code words, shared references, local slang and inside jokes (not that Personal Vocabularies are actually available just yet, you understand).

Going from there.

Arranging the (64) related terms in each "Created Division" of the description layer and creating adjectival and adverbial variations is the hard part (or at least the manual part). Generation of plurals and degree forms, on the other hand, is mostly trivial. Picking the 'common' (#00) term for the "Neutral Description" can be simplified with usage counts.

Unfortunately the relationships between the terms and the Types are much less self evident at this level and index collisions are bound to occur. The problem is that a large number of these decisions are essentially judgment calls. Generating action and entity nouns and determining which nouns are "Non-Descriptive" is hard to delegate to a machine. These processes can be semi-automated by using dictionary checks, but no dictionary can ever provide one hundred percent coverage.

Homonyms will need to be identified and clarified (bark as in dog or bark as in tree?). Verbs will also need to be nouned (nounified? ... noun-o-matized?). Most verbs can be treated as nouns without modification (a run, a look), but there always suspicious transients loitering around (a see? ['scene' {entity} and 'sight' {action} are the more usual suspects]).

This is where an active involved user base would actually come in handy. "With enough eyeballs all meaning is shallow." The application needs an integral 'feedback' button for inarguable fixes. The feedback button would automatically create a transaction that could propose a new (or changed) term for a given location. A "Swap Multiple Terms" button would be very handy also (no idea how it would work just yet).

Of course, a feedback button won't even be practical until after a vocabulary is computer generated and the button wouldn't be available on the free version of the application. We need to limit "easy updates" to people who are serious about the results. If you actually pay to use something you are less likely to try to screw it up for vandalistic ego boo.

The Specialization Layer

There is a list of Specialties having Focal vocabularies. These Specialties could be distributed using the same process as the basic terms. This is only the first step to actually creating the Specialization Layer, but it would (hopefully) be a good place to start.

Collecting all of the specialized terms is not difficult, but arranging them logically could prove to be the work of several lifetimes.

back to the home page.