3.5 Reasoning in natural language
By implementing upper ontology1 techniques in natural language, the system is be able to reason in natural language: generating new knowledge, asking questions and detecting conflicts.
Domain ontology will be implemented as text files containing information about certain subjects, written in natural language.
Conclusions
An example of an automatically generated conclusion: Given the sentence “John is the father of Paul.” will result in: “Paul has a father (called John).”.
Assumptions
Given: “A person is a man or a woman.” and “John is a man.”. The conclusion “John is a person.” can not be justified. So, it is called an assumption, expressing the uncertainty.
Questions
Given: “A parent is a father or a mother.” and “John is a parent.”. Without further knowledge about the gender of John, the system will conclude: “John is a father or a mother.”. However, in order to encourage the user to complete the knowledge, the conclusion will be converted to a question: “Is John a father or a mother?”.
Conflicts
Given: “A parent is a father or a mother.” and “John is a father.”. When a sentence is entered like “John is the mother of Paul.”, it will not be accepted, because it is in conflict with “John is a father.”.
1 also known as top-level ontology or foundation ontology