Tag: artificial intelligence

Artificial Intelligence –
Semantic Networks

Artificial Intelligence –
Semantic Networks

semantic network (or frame network) is a network that  represents semantic relations between concepts.

They had been designed to represent the meaning of words in natural languages. The information is represented as a set of connected nodes, with labels along the edges showing relations among nodes.

They are strictly linked with the logic of the predicates: in fact, the relations in the logic of the predicated are now represented with states.

Inheritance Algorithm I (in pseudo-code)

This algorithm is used to get the value V of an attribute A of an examplar of O.

  1. Search for Object O in the knowledge base.
  2. If O has an attribute A, get the value V!
  3. Else
    1. if O has not an attribute A type exemplar-of: end with fail.
  4. Else move to the node with value associated to examplar-of attribute and go to point 2.
Problem of this algorithm
  • It’s possible to have more than one solution.
  • It is not possible detect ambiguity of the representation.

Inheritance Algorithm II (in pseudo-code)

To reach this algorithm a new concept has been introduced: the inferential distance. It has been defined by Touretzky (1986) in this way:

“Class 1 is nearer to Class 2 with respect to Class 3 if Class 1 has a inferential path that ends to Class 3 moving across Class 2”.

To get the value V for an object O for an examplar E:

  1. Put CANDIDATES as empty.
  2. Execute a breadth-first search or depth-first search starting from E, towards the top, following all the edges examplar-of and isa. At each step of the process we look at the value
    1. if the value is there, we can add it to CANDIDATES.
    2. otherwise if there are edges available follow and go point 2.1
    3. otherwise: end the edge.
  3. For each element X of CANDIDATES we look at the inferential distance of X and compare this with the others to find a nearer X2.
    1. if found some lower inferential distance value, remove C.
  4. We check the length of CANDIDATES
    1. if 0: no value found.
    2. if 1: show the value.
    3. if >1: there is a contradiction.