Andrey Kuryan, Mikhail Rubin: General information about the project. Ontology “TRIZ Model”.
https://triz-summit.ru/confer/tds-2020/web/1/ (in Russian).
Hans-Gert Gräbe :
Владислав Ерёмин : Где можно ознакомиться с актуальной версией словаря?
Ramez Kassou : Идет ли кем-то проработка новых ценностей, которой дает построение онтологии ТРИЗ? Такая большая работа только для ради системного образования? А как насчет интеграции с системами ИИ например?
Hans-Gert Gräbe :
Ramez Kassou :
Hans-Gert Gräbe : Как вы справитесь с противоречиями? Общие позиции - это еще не все.
Александр Быстрицкий : Обучение - это процедура формирование согласия.
Ramez Kassou : Спасибо большое за приглашение, за вебинар и за большую работу!
Артамонова Анна : Спасибо большое за приглашение, очень интересные были доклады и дискуссии!
Александр Быстрицкий : Это очень интересно, то что сделано заслуживает и внимания и уважения. Интересно поработать совместно, хотя разница в подходах школ весьма существенна.
Hans-Gert Gräbe :
Vladislav Yeryomin : Where can I find the current version of the dictionary?
Ramez Kassou : Does someone work out the new values, which gives us building a TRIZ ontology? Such a big job just for the sake of systematization? What about integration with AI systems, for example?
Hans-Gert Gräbe :
Ramez Kassou :
Hans-Gert Gräbe : How will you manage conflicts? Common agreement is not all.
Alexander Bystritsky: Training is a consent formation procedure.
Ramez Kassou : Thank you so much for the invitation, for the webinar and for your hard work!
Anna Artamonova : Thank you so much for the invitation, there were very interesting reports and discussions!
Alexander Bystritsky: It’s very interesting, what has been done deserves attention and respect. It is interesting to work together, although there are very significant differences in the approaches of the different schools.
(Addendum and additional remark by Hans-Gert Gräbe)
Most of our computer science students do not even understand the difference between a data model and a database scheme, but this is important, and there are plenty of tools to transform a data model into a (even normalized) data base scheme (at design time) and also much tools (e.g. the Hibernate framework) to perform such a transformation at runtime. This allows to work with data abstraction layers (e.g. DAO) within the implementation and use a general and a well defined persistence layer to organize all the storage processes.
An ontology is such a data model that is widely accepted and can be used by integrated tools as, e.g., the persistence layer framework within the technology just mentioned. Hence an ontology is a socially accepted standard and a social process has to be passed through successfully.
What is the core of such a process. Take any TRIZ notion, e.g. “TRIZ methods”. It is a concept, but for the computer it is solely a wording without any semantic meaning. The computer only understands, that the same wording at different places means the same. Humans act differently, since they use at the one hand in some cases the same wording for different meanings and in other cases different wordings mean the same. NLP and ML is about that - to map wordings on unique concepts (since the computer hardly can handle fuzzy notions). Humans behave differently in another concern: they can good live with notions that mean almost the same. For computer use this has to be “reduced to the essential”, i.e. the “almost” has to be removed putting the whole situation in a restricted contextualization (details are explained in my TFC 2020 paper).
This (social!) process is called formalization of the ontology and can be done in a very informal way using plain text. To make the interchange more machine readable the language itself, that is used for the description of the model, has to be formalized in a metamodel.
This applies in particular the arrow labels in the model as “is related to” (irt for short). If you have two such relations, as e.g. irt(RTVModel,TRIZModel) and irt(ProjectActivity,TRIZModel) in the “TRIZ Model Ontocard” (p. 9 of the TDS 2020 paper) one has to understand, what “is related to” means, in particular, does it mean the same in both contexts.
First, I wrote the predicate irt as function with two arguments replacing infix by prefix notation. Secondly, this are signatures, in the brackets the argument types are listed. In OO programming there is no problem to have the same function name with different implementations. The correct function (i.e. the appropriate pointer to the code) is chosen according to the type of the arguments. Thus using the same function name a hidden unification process is present that has to be made explicit on the level of the metamodel. This, in its turn, can be done informally using plain text to explain to other humans what the computer is doing.
But if you plan to automate also that process you need a language for the language description, hence a metametamodel. These three levels are nowadays more or less standard in ontology modeling.
In many applications RDFS https://www.w3.org/TR/rdf-schema/ is used as metameta model, since it turned out that OWL https://www.w3.org/OWL/ is much to complicated for that (if the model is complex enough, it can be strictly proven that OWL reasoning is undecidable). In many ontology modeling applications nowadays SKOS https://www.w3.org/TR/skos-reference/ is used as such a metametamodel, although they claim “SKOS is not a formal knowledge representation language” (cited from the website). This is indeed true, hence a metametametamodel is required (in this case RDFS is used).
Hans-Gert Gräbe, last update Oct. 21, 2020