RSS AI Zone
Yeah Dave, I’ve studied grammar and languages and linguistics extensively, I’m aware of the standard definition of “verb” 
What I’m after in my system is the relation, so I can put it into a graph as an edge. For example:
John ate an apple
would be stored in the graph as a Vertex “John”, a relation (or edge) “ate”, and another vertex “an apple”.
1 kg is heavier than 1 g
would be stored in the graph as a Vertex “1 kg”, a relation or edge “is heavier than”, and another vertex “1 g”.
Loosely speaking the vertices are nouns or noun phrases, and the edges are verbs (but can include modifiers and other grammatical elements).
An example of a graph with the kinds of relations I’m describing might be: http://hyperphysics.phy-astr.gsu.edu/hbase/hph.html#mechcon. This graph is not completely consistent in its labeling of edges, but it contains a few edges labeled with verbs + other elements such as prepositions, and the vertices are often objects of prepositional phrases.
The advantage of such a non-conventional approach to chunking sentences is that you can retrieve knowledge from it through questioning easier. So:
Is 1 g heavier than 10 g?
will search in the graph for the vertex “1 g”, and then look for an edge labeled “is heavier than” leading from that vertex. If it finds such a relation it returns the contents of the ajacent vertex; if not it returns “no”.
