Title: Modeling Questions
Description: This problem came up in the context of the W3C Decision Incubator activity, however, it is more general than the decision-making domain. It is about how to formally represent questions in OWL.
For instance, if I ask a question like "Where did this emergency occur?" I could simply represent it as a string. However, the question contains a lot of information, so in some cases one might want to represent it formally, and possibly reason on the question itself. For instance, we may want to derive that the answer should be of type "location" from the fact that the question uses the keyword "where".
We have identified two main aspects of this modeling issue:
- How can we refer to the detailed description of the question?
- How can we model the semantics of the question itself?
The first issue concerns the case when we would like to include both the question as a string, i.e., the human readable version of the question, and the question as a formal model. In this case we may have a class "Question", with a string property containing its representation in natural language. How do we now link also a formal description of this question, i.e., a set of triples, to the question class?
The second issue concerns how to represent the semantics of the question itself, i.e., an ontology for modeling those triples we wanted to refer to above. We would like to model things such as, the variable of the question, the expected answer type, any other constraints set in the question etc. For instance, in order to be able to ask questions such as: What is the expected answer type of this question? Is this answer correct with respect to the constraints of this question?
There exist some question classifications originating in expert systems and question answering, however, we are not aware of any OWL models.
Diagram (this article has no graphical representation)
|Competency Questions||What is the variable of this question? What are the additional constraints set by this question? What is the type of thing this question is asking for?|
|Scenarios|| In the decision-making context, I may be about to decide from what shop to buy a certain product. My question could be "From what shop should I buy a car of model x, manufactured later than 2003?". My decision will be the answer to that question. In this case the variable is indicated by the "what" keyword, and restricted to take values of type "shop", or possibly even more specifically "car dealers". The model an year restricts the answer set for the question, i.e., I don't want the answer to be some car dealer that doesn't sell model x. By representing this question formally, rather than just as a string, I would be able to relate the parts of the question to things such as metrics for assessing my options (i.e., the different shops), and checking that a proposed answer is actually of type "car dealer" and "shop".
Another scenario might be in the domain of emergency management and there is a need to transport a patient to an appropriate medical facility; however, in an emergency, some facilities may be overloaded and may not have the needed services. So the question might be "To which medical facility should I send this patient X with needed services Y?" If I represent this question as a string, it is understandable to humans but not to machines. How do I represent this question in OWL to answer certain competency questions such as: What is the Subject of this question? What is the unknown part of the question we are trying to answer? Is it the subject, the property or the value? Is the unknown part a who, a what, a where, a how or a why?
|Proposed Solutions (OWL files)|
|Submission to event|