Basic Categories

Consider the following sentence:

"Smart people contemplate."

and its SNePS representation:

When the verb algorithm is run it will respond with the result:

A SOMETHING CAN contemplate with result() or enable ()

It fails to report a more detailed description of the agent because the verb algorithm can only report two things:

1. Anything that is pointed to by an agent, object or indirect object arc and is also modeled as being basic category.

2. Anything that is pointed to by an agent, object or indirect object arc and is of class animal. In which case it is reported as being an animal.

All other instances are reported as being "SOMETHING"

This is somewhat disconcerting because there is a large difference between knowing a smart person can contemplate as opposed to something.

I suspect that what you would want to do is this:

1. Find the act, object or indirect object arcs, then either:

2A - Have the verb algorithm examine the structure and return an appropriate english translation. This seems quite unlikely as the algorithm would have to be prepared to handle an infinite number of case frames,

2B. Ask whatever is at the end of these arcs to give back an english translation of what it represents. This might require each item that can be at the end of an act/object/indobj arc implement a rule that returns its english translation. The algorithm could search on:

ANT ! CQ ENGLISH_TRANSLATION

I think that 2B is a step in the right direction but I had difficulty envisioning how it would work. Essentially how can you create a rule that will have a case frame return an english definition if it itself is composed of other case frames?