]Analysis: 1. level
Interpretation-generator
Preconditions for using Interpretation-Generator:
- The results from the similarity analyses are visualized in two tables: table of stairs and table of estimations.
- The table of stairs is the primary result itself, because the model derives those stair functions (stair-levels per attribute) from the learning pattern that fits most to the given restrictions. The stairs represent the automatically generated knowledge that allows computers to determine which consequence fits most to the input variations that are admitted to be unique. The knowledge of the stairs may overhang the domain of the known samples, so such consequences can be drawn that occurred never before (cf. genetic potential). The stairs show the importance, sensibility of each attribute, and their ceteris paribus and polynomial effect, the existence of innovational signs. The interpretation of the stairs are partially context free, so they do not depend on the name of attributes, but partially they are context-dependent, so it can give a basis of associations to the directions of those knowledge elements that are not included within the OAM that represents the learning pattern.
- The estimation table is some kind of secondary view of summary: it is nothing else than the projection of stairs to the objects that represent the learning pattern. So the estimation table exchanges the primary and/or ranking data with the proper stairs, and gives the estimation value of the objects as a resultant. The deviation between estimations and the facts is the model error. The deviation between estimations and facts may be interpreted as either under- or overrated (loss of balance), or as brand value, or the value of missing data.
- The interpretation-generating (cf. expert system) is nothing else than the choice of the proper template based on the results gained (relying on a template collection that was created intuitively based on previous experience).
- The intuitive template generating is nothing else than recognizing the logic that lies in similarity analysis step by step (keeping Vonnegut's Hocus-Pocus problem in mind).
- The automation of interpretation enables automated generation of textual reports. It is not equal with the formation of those driving signs that are directly based on the analyses (e.g. automated omission of direct antagonisms from the OAM upon revelation). The textual reports serve as the support of humane decision-making and the lingual handling of problems so they rouse further associations.
- The initialization (planning) of models is not to be confused with the direct interpretation of the results. Because the analyses makes sense in the analysts' head as part of a higher relationship, the automated initialization of new runs that are dependant on the previous ones is possible only after a structured description is given to the problem on the holistic level. But it is possible at least!
- Robot-experts may be created at any point of the automation tasks: thus, the simulation frameworks that work with stairs, the handling of textual report making and the model chain that are based on each other are all regarded as robot-experts. Thus if something was comprehended by the human brain intuitively, then it may possible to handle that thing with a computer too...
