Each Concept Network link "belongs" to a node, from which it comes. Its is directed to another node, and has following fields:
- type
- contains the name of the link's type (contains, co-occurs with, followed by, etc.)
- node
- the node pointed by this link
- weight
- the influence that the departure node has on the arrival node, that one can also see as a conceptual proximity of the two related nodes.
- label
- optional pointer to a label node, turning the link into a labeled link
The next figure shows an a minimal Concept Network sample, where the birthday node is completely activated (100%). This node is linked to cake, whose activation value is zero. This link's type is eat, and is weight 95%. This network is an abstract representation of the fact that for an anniversary, one often eats a cake.
/----------------\ /--------------\
| S: birthday | | S: cake |
+----------------+ +--------------+
| CI: 80 | /-----------------\ | CI: 70 |
| AV: 100 % +-----+ T: eat | W: 95% +---->+ AV: 0% |
| DR: 2 | \-----------------/ | DR: 5 |
| Ag: ComputeDay | | Ag: CookCake |
\----------------/ \--------------/
Like in neural networks, the activation value from birthday will be propagated to cake, thanks to the activation propagation mechanism.