A Generative Approach to Variations in Production Processes.
PierPaolo Brotzu, Marco Maiocchi – TEAnO – Milano

Many work procedures in generating various types of products (such as execution of recipes for food, work operation for leather, procedures for conservation of vegetables, factory workmanship processes, etc) are characterised by precise procedural steps.
Each step can be defined, at the proper detail level, in terms of kind of operations, input conditions, output results; then, a complete procedure can be exploited as a set of partially ordered steps of such a kind of functions.
The formalisation of such a steps allows to evaluate procedures devoted to the same goal, introducing metrics for comparing them in terms of originality or imitation.
Furtherly, a contextual analysis of different procedures devoted to similar goals allows to verify possible variations in the procedures, introducing innovation and changes on the basis of the available examples. The changes can be automated through proper algorithms, just using techniques of learning by examples, in the respect of verifiable constraints.
The results can be verified, and used again as a new enriched experienced basis to improve the innovation procedures.
The paper will describe the approach for the analysis and the description, will propose an experienced metric, and will show examples of generation of new innovative methods based on the automated analysis of successful examples.
All the practical examples will be expressed in cooking recipes, described, compared in terms of plagiarism with properly defined metrics, improved with automatic generation based on experience and examples, properly checked and validated.