author = "Simon Parsons",
title = "Current Approaches to Handling Imperfect Information in Data and Knowledge Bases",
journal = "Knowledge and Data Engineering",
volume = "8",
number = "3",
pages = "353--372",
year = "1996",
url = "citeseer.ist.psu.edu/simon96current.html" }
** Classification of imperfections in information:
*Category && Sources
uncertainty: arises from the lack of information about the real world.
imprecision: arises from a lack of granularity.
imcompleteness: a lack of relevant information
inconsistency: arises from having too much information from too many sources
ignorance: used to describe a lack of knowledge.
**Imprecision && uncertainty
type1: "His age is between 25 to 30" -> interval valued
type2: "he is quite young" -> fuzzy valued
type3: "he is either 26,27 or 28" -> descrete form of imprecision.
type4: "he is not married to Ann" -> imprecision from negation.
Imprecision is thought as objective and arising from the granularity of the language used to make the imprecise statements.
e.g. "he is 26 years old" -> might be a precise statement only if we are not interested in he's exact age in terms of years and months.
>>usually imprecise information is modelled by "fuzzy set" and "fuzzy logic".
uncertainty is an estimate of the truth of some fact by some individual(subjective). it arises from a lack of information about the state of the world. This lack of information makes it impossible to determine if certain statements about the world are true or false - all that can be done is to estimate the tendency of the statement to be true of false ...
>>in DB area: approaches based on possibility && probability theories
>>in AI area: logical approaches (possibility logic, probabilistic logics) V.S. Numerical approaches (Bayesian belief networks)
*uncertainty and imprecision may arise together in the same piece of information.