We developed a system that can perform predicate logic operations similar to PROLOG based on natural language. We then made it possible to express and reason about "strong negation" (taboo expressions) that is morally or scientifically forbidden. This is different from weak negation, i.e., the impossibility of proving in the closed world of PROLOG, and was very useful for describing ethical taboo knowledge that constitutes big data, but it did not support exceptions to taboos or taboos of those exceptions. In this paper, we propose a knowledge representation method that can simply describe the priority switching and logical integration of the "strong negation" (level 1), affirmation as the "stronger" exception (level 2), and negation as the much "stronger" exception(level 3) of that exception (level 2), etc., with priority. We implemented the improved system that can be interpreted and inferred by a computer, and proved its practical use through simple experiments.

Traditional logic programs are written in some formal languages easy to “unify” each other or symbol based ones such as in Prolog, and their “NOT” mechanisms are “negation by failure of proof” according to “CLOSED world assumption” and are not assumed as “true negation”s.On the other hand, we suppose that BigData for moral, ethics and values are written originally in various different natural languages over the world.In order for both human and AI to easily understand and automatically operate such BigData, we hope that (1) human could logic-program the contents of such Big-Data, by semi-dead-copying the original-natural-language-texts as "logic-programs" without hard translation, formalization and/or normalization,(2) “NOT”s in the language text that mean some TABOOs in any culture shall be operated correctly as such (not as “negataion by failure of proof”s but as the TABOOs) even in case of concretizing objects from different cultures, as in OPEN world assumption, because any TABOO from any one culture shall be rigidly respected, even during meetings between people from different cultures including the one culture with the TABOO.For the above purpose, this document explains a method of natural-language-based-logic-programming where variables for knowledge abstraction and concretization are embedded into the natural language texts, and also explains the above strong "TABOO-NOT" implementation therefor.

If there is a format that is easy for both humans and AI (computers) to understand and infer, and also if the humans let the AI learn various morals, ethics and values in the above format, these data may be operated automatically by the AI in a way that the humans concerned are able to check for mistakes / lacks of thoughts in the automatic judgments by the AI, avoiding arbitrary thoughts / judgments by both the providers(humans) of such data and the AI.On the other hand, morals, etc. are diverse depending on the country, history, culture, social position, and/or individual. So in order to accurately express such data for reducing interpretation mistakes, lacks of required conditions of thought-rules and so on, we need to allow, as such data, expressions almost in (or from completely including all) the natural languages used in each cultural area.In this paper, we (1) propose a format based on natural language that we called as "situation-specific-value-order-formula" as a starting point for an expression format that satisfy the above needs, (2) explain its components: situations / viewpoints and items ( positive / 0 / negative ) ordered by the value all based on natural language, (3) explain related things including the specific l ogical i nference f eature o f t he n atural language based programming where variables are embedded in the natural language.We conclude that our proposed format is feasible and practical to some extent for the above Purpose and need.