Brain Theories, Applications 1
Language, Semantic Nets

An immediate problem with the brain model as described is that links seems to need to have attached properties. How is this to be modelled? For example, consider Langacker's semantic networks model. The basic scheme involves groups of items to be interconnected, just as would be allowed by the present brain model. However, a distinction must be made in Langacker's scheme between different kinds of collections, that is the base of a concept involves a connection to a wide group of other concepts, while the profile involves connections to a narrower set, usually a subset of the base group. Each group of interconnections must be labelled in some way which is accessible to (i.e. particpates in) the concept being defined.

There are a couple of ways this labelling could be done. In terms of the brain model, the problem could be stated as a requirement for simultaneous, overlapping activated patterns. One way of doing this was discussed in the brain model essay, that is, the use of separate oscillator frequencies. Each group type could be connected by firing patterns with different frequencies. The base group would connect all of the relevant base-related concepts, together with whatever other circuitry implemented the base pattern type. The profile group would be connected with the smaller profile-related concepts and the profile pattern type circuits.

A second way to maintain independent, overlapping groups would be by the use of separate node types, thus implementing separate networks which would not be crossconnected. One sees a distinct disadvantage of this method in the lack of ability to crossconnect networks. For example, the circle and arc concepts described by Langacker require cross connections in that the base of one is the profile of the other.


Show list of book reviews sorted:
by Topic | by Title | by Author
Show list of essays sorted:
by Topic