Meaning Operators

We are using meaning operators in two ways – either as complete or partial branching.

Complete branching:

The word “run” is either a noun or a verb – there are no other possibilities.

While we wait to determine which it is, we connect to the word, then move the connection to a particular part of speech child.

Partial branching for “die”:

The head node is general, the child nodes are specific cases, which cover perhaps 10% of the meaning of “die”.

When we are sitting at the head node, there is no indication that we have discounted the applicability of the child nodes.

We could handle this by introducing a state for the connection – temporary or permanent – or introduce an additional child node, which represents everything else but the existing child nodes – let’s call it “NotOtherChildren”.

That is, in this case, it is a device for cutting or molding of metal which is not an engraved device or a device for cutting threads. This would be much more meaningful, but would cost tens of thousands of network elements.

And it still wouldn’t be correct. An “injection molding die” is a device for molding plastic using injection – it may be correct to call it a “mold”, but an “injection molding mold” sounds silly.

How often will we need this approach – not all the time, as some subsenses should be folded back into senses, and we will treat the sub senses for prepositions as exhaustive. While expensive, it opens a door to knowledge expansion – if we can show something meets the requirements of the head but is not any of the other children, we attach to NotOtherChildren and wait until more information becomes available (if the node is not there, we create it).

Is this good enough? We might know it is a die for plastic, not metal, and does not match any other child on this level – there is no reason why NotOtherChildren cannot start sprouting children and become a normal part of network traversal.

This may be a simpler example:

     Bang
     Sense 2
      Definition: make a sudden loud noise, typically repeatedly
      Example count: 1
       Example 0: the shutter was banging in the wind
     Sub Sense 0
      Definition: (with reference to a door) open or close violently and noisily
      Example count: 2
       Example 0: he banged the kitchen door shut behind him
     Sub Sense 1
     Definition: NotOtherChildren  (that is, not a door)

If it is a window banging, then we know it is not Sub Sense 0 (it probably shouldn’t be a sub sense, as it is not “typically repeatedly”).

An example of an extreme case of a sub sense:

Bar
Sense 5
    Definition: a barrier or restriction to an action or advance
    Example count: 1
     Example 0: political differences are not necessarily a bar to a good relationship
    Sub Sense count: 1
     Sub Sense 0
      Definition: a plea arresting an action or claim in a law case.

Here the sub sense would apply in perhaps 1% of instances, and could easily be ruled out.

Meaning Connections

One meaning of “to croak” is “to die”.

One approach would be to connect the particular definition of ToCroak to the particular definition of ToDie through an assignment (one way) operator, as shown.

If the connection included relation operators, the downstream elements (the words used in the definitions) would be children of the associated objects, so the dictionary entries and their definitions remain reasonably clean (not have thousands of connection to other nodes, which are doing all sorts of other things). It seems reasonable to do the same for single connections, so the connection between ToCroak and ToDie becomes as shown.

It has a problem, in that the particular definition for "to die" applies to a person, animal or plant, so the machine could assume a plant can croak.

The 2,336 uses of "with" in the definitions will now appear as instantiations below the line.

This approach allows separation between the dictionary entries and the steadily growing connections (potentially millions) resulting from, first the definitions, and then the use of the words in text.

The continuing desire to save a few network elements leads to poor outcomes.