[AZ]

There are a many generators for a grammatical structure of text (basing on, say, the control by verbal forms and adverbs, collateral subordinations of sentences, etc.). But I am not sure that the knowing of such grammatical structure of, say, patent texts helps to state what a patent is more preferable to be used.

The semantics of a patent text which is most interesting for a customer.

In one word, you are absolutely right: there are a lot of deep problems here. We need a very investigation in order to understand (and then to formulate and to formalize) what a sense of text means when compared to a real natural world, and how that sense depends on the ultimate goals of a customer, etc.

Now I can only mention once more as to my experience based on VISAD-technology and related to the extracting a very non-trivial sense from the text.

I created a text database on known logical and set-theorerical paradoxes. Every paradox is a text. Every text was decribed by a set of characteristics (about 30) concerning any properties of the texts such as a length, a number of sentences, nouns, verbs, adverbs, grammatical properties of this terms and connections between them, etc.

It was important that the set be generated by an expert group (3-4 experts) in an interactive regime basing on a visual voting method Anton, please send a concise description of this voting method..

Then an automated classification was realized and the VISAD offered to experts its solution. After some iterations "classification-expert evaluation-classification" we get a semanitcs of the texts set in the form of the explicit set of necessary conditions of the paradoxicality. Anton, why can we not simply write up this experiment so as to close the current contract?.

Our work allows us to formulate the main semantic result in the form: NOT the SELF-REFERENCE as such (according to Russell) but the SELF-REFERENCE together with the NEGATION (the NEGATIVE SELF-REFERENCE) is a real necessary condition of the paradoxicality. (Note by Paul Prueitt: " Thus we must include the type of negative information as most everyone in Russian semiotics community (Pospelov, Finn, etc) agrees.")