The physical representation of semantic space

Citation

James, B. (2001), "The physical representation of semantic space", Kybernetes, Vol. 30 No. 1. https://doi.org/10.1108/k.2001.06730aaf.003

The physical representation of semantic space

The physical representation of semantic space

Keyword: Cybernetics

Abstract Discusses the concept of semantic space and its relationship to the objective and subjective universes which we recognise. Posits a method of representing semantic space physically in relation to mass, space and time.

We can recognise three universes: the objective universe; the subjective universe, which is what we see; and the semantic universe in which we express our thoughts. Darwinism has made the first two resemble each other closely, giving an obvious advantage, but clearly the sky is not blue and the grass is not green – those are just two subjective sensations present in the subjective universe. One can think of an idea, then another similar but distinct from it then another or another between two of the previous ones just as one can consider numbers. But numbers expire at infinity, so are there margins to semantic spice? It would appear that space time and mass are sui generis so, provided that consciousness can be adequately described in terms of mass, space and time, semantic space is a triangle with mass, space and time at its apices.

If consciousness cannot be explained in terms or mass, space and time, semantic space will be a tetrahedron with m, l, t and c at its apices with ideas represented as points in such a way that similar ideas will be closer than disimilar ones, which one could call a Roget map.

Most of the objects one can see are arranged almost at random, so one of the first tasks is one of classification. Some of this has been done by Darwinism so young children recognise faces and living things by the fact that they move, but the advance in human thought has improved the manner in which we recognise and classify objects. In semantic space we can describe mathematics as that part which recognises the advantage of the use of number as an adjective, and science as that part which relies on experiment for verification. Problems are the determination of the route between the problem point A in semantic space and the solution point B. Since there are initially no landmarks in semantic space the first solution will be found by a drunkard's walk. In practice one solves problems by the application of logic to factual information, but since both the logic and the factual information were determined initially by trial and error the principle, which derives from the second law, is preserved.

In a reductionist universe everything eventually will be explained by science, so all semantic space will be reduced to a network of inference chains which in principle could be checked by experiment. The first step in science is the botanical phase where everything is named and classified. This means that every signal must be memorised and the similarity of new signals to old ones recognised. The main input to the brain is vision, which usually consists of a moving object in the foreground against a relatively stationary background, and one of the first tasks performed in the retina is to separate the moving images from the static ones.

In the triangle of semantic space we can draw a line between I and t – points to the I side will be static objects, nouns and points, while to the t side will be moving objects and verbs, with the distinction being made in the retina. Nouns and verbs are often associated with each other – nouns can be verbed and verbs nouned; for example, tabling an amendment refers to the table on which the document is placed. Conscious thought is probably confined to the cerebral cortex, but it is not one semantic space but many, the problem being made more difficult since it is folded in a manner which is random to a good extent so the surface must be normalised for accurate mapping to be performed. Digital maps of the cortex have been published with a resolution of 0.3mm.

The unfolding can be accomplished in the following manner. Convolutions are on the surface – it is required to bring up the sulci. Choosing two fairly adjacent points on adjoining convolutions we ask the software to determine the shortest route between them, and this will become a great circle in the isometric cortex. Choosing new points, the process is continued by triangulation until the whole surface has been normalised. The new shape has been determined roughly by pencil and paper, and is relatively simple but require accurate determination by software for verification. The distribution appears to be as follows.

Space is contralateral and inverted, specialised functions to the rear and generalised functions to the front, verbs (sets in time) laterally and nouns (sets in space) medially, each space being organised as a Roget map. There may be as many as a thousand cortical areas and each maps onto ten others, so three iterations will paint the entire area.

Brennig JamesCherry Orchard, Marlow Common, Buckinghamshire, UK