Measuring the Quality of Uncertain Information Using Possibilistic Logic

Abstract

Abstract. In previous papers, we have presented a framework for merging structured information in XML involving uncertainty in the form of probabilities, degrees of beliefs and necessity measures [HL04,HL05a,HL05b]. In this paper, we focus on the quality of uncertain information before merging. We first provide two definitions for measuring information quality of individually inconsistent possibilistic XML documents, and they complement the commonly used concept of inconsistency degree. These definitions enable us to identify if an XML document is of good or lower quality when it is inconsistent, as well as enable us to differentiate between documents that have the same degree of inconsistency. We then propose a more general method to measure the quality of an inconsistent possibilistic XML document in terms of a pair of coherence measures.

A knowledge-based approach to merging information

“…In other papers, we have (1) presented an outline of using for fusion rules for knowledgebased merging of structured reports [Hun02a]; (2) presented a range of aggregation functions for use in fusion rules [HS04a]; (3) presented a framework for using temporal logic in knowledgebased merging [Hun02c,HS05]; (4) explored properties of a restricted form of fusion rules [HS03b]; (5) developed a framework for measuring degree and significance of inconsistencies in information in order to choose how to act on inconsistency with actions including ignore, resolve and reject [Hun03,Hun05]; and (6) developed aggregation predicates for merging uncertain information [HL05a,HL05b,HL05c,HL05d]. This paper extends the previous papers by providing a more general framework for knowledgebase merging and by providing comprehensive experiential insights into practical aspects of developing knowledgebased merging using fusion rules.…”

“…To deal with these situations, in [HL05b,HL05c,HL05d], we have further extended the approach to merging multiple pieces of uncertain information to situations where • evidence is specified at different levels of granularity on the same concept as textentries. We refer to two pieces of this type of evidence as semantically homogeneous.…”

“…Using Dempster-Shafer theory, we can harness Dempster's rule of combination that allows for the combination of two or more uncertainty distributions [HL05a,HL05b,HL05c,HL05d]. This forms the basis of a useful class of aggregation predicates with intuitive properties.…”