Units of measurement in semantic space. Measure of distance in semantic space.
Traditional and widespread unit of information measurement is 1 bit. Since information is a kind of object condition measurement, then differences between objects as representatives of its classes also could be expressed quantitatively in bits.
It is offered to use a vector, which determines difference of co-ordinates in notion space.
The sense of this formulation could be described as “what information should be changed in notion N1 for turn it into N2.”
Obviously, that |Si| is that distance between notions in bites. Offered name – Shennons.
1bit=abs (1 shennon) – minimal distance between notions corresponded to 1 bit difference in semantic space.
But such easy and obvious theoretical approach doesn’t exclude practical complexity of this problem. Before attempt to calculate in practice those distances between notions followed problems should solved:
· Problem of language. Determinations of notions in languages don’t correlated directly (linear and unequivocally) to the notions themselves. European languages are non-optimal code for notions. They bring serious distortions in notion’s representation. Hence, to calculate co-ordinates of notions with information inherent in European languages is almost impossible.
· Problem of personal perception. Personal semantic space exists for each human. And it isn’t similar to common notions. That is why it is impossible to construct the semantic space with help of only one person or small group of persons. Obviously, that “objective” picture of semantic space is able to give only group of persons representative enough, this means several hundreds of people. Hypothetically, if specialists in the field of language (linguists, literary men, philologists, teachers of languages) had been attracted, the volume of extract could be decreased.With using of fractal model the approach to measuring of distance between notions is interesting. Then this distance would be detecting with арность of graph, which includes both measured notions.