June 27, 2007 1:58 AM
Consensus on semantic data and procedures
In this paper "Consensus Data, Ontologies and Services" Prof. Michael N. Huhns has presented examples of what can be achieved with semantic tools. He emphasizes on consensus in order to form a service model that can solve the problem in a given domain. He suggests in order to achieve result what must be specified. Here we will discuss some examples from this paper and address the other part of the problem, how this can be achieved.
Example 1: "What is the population of China?" Here the issues are the correct interpretation of the query and authenticity of the result. Assuming normative metadata, RDF web service providers will have this information in the form of triples:
country population _:pvalue
_:pvalue year "value"
_:pvalue total "value"
_:pvalue men "value"
_:pvalue women "value"
_:pvalue children "value"
If the year is not specified in the query, the agent will assume current year to be the implicit value of the year. The agent may thus return the results from web service providers that have the total population value for the current year. The authenticity of the data may be determined by the agent based on the internet domain address of the data IRI (e.g. .gov, .edu, etc.) or left to the discrimination of the web user.Example 2: "If current trends continue, when will the population of India equal the population of China?" There are multiple questions in this query that must be answered, this requires converting this query to an appropriate SPARQL query. To answer "if current trends continue", the current trend must be found and then a consensus has to be built to find the most appropriate value of the trend. The agent may aggregate the trends from the authentic providers and present the query answer along with the IRIs of data providers.
Example 4: "How can two sources each having semantics specified independently be reconciled or enabled to interact without confusion?" Here instead of consensus on semantics the use of normative metadata will enable interaction without confusion through context mediator.
Example 6: "What is the best, or consensus, recipe for baking chocolate chip cookies?" as the idiom "too many cooks spoil the broth", mixing recipes will not bake the cookies. The solution suggested is to either find the recipe IRI through the chocolate chip cookie product RFID URI, or combine the marketing research web service provider results to find the most popular chocolate chip cookie and return the recipe IRI.
To conclude: In this study of examples it is found that instead of consensus on semantics it is a better idea to use normative metadata for authenticity of information. Consensus may be required when more than one web service providers have different values for the normative metadata. The objective is to derive to-the-point answer to the user query in order to reduce the effort required by the web user to read the million order IRIs returned by the search engine like Google.
I am going to be blogging live from a couple of days of the 
Leave a comment