作者: Thomas Lukasiewicz , Rafael Peñaloza , Ismail Ilkan Ceylan
DOI:
关键词:
摘要: Conjunctive query (CQ) answering is an important reasoning task in description logics (DLs). Its goal to retrieve the tuples of individuals that satisfy a conjunctive query; i.e., finite set atomic queries. These are called answers. Clearly, given CQ may have considerable number answers, specially if individual names appearing ABox large, as case for many existing DL ontologies. In order manage all these answers structural manner, one can try extend with preference criteria, such way most preferred returned first. Possibilistic networks (PNs) arisen representing conditional preferences over events compact [1]. The general idea provide possibility degree each event which proportional event. We apply this model indirectly, by modeling contexts entail them. nutshell, we divide EL knowledge base (KB) into contexts, and use possibilistic network describe joint distribution contexts. Our formalism based on ideas previously presented under probabilistic uncertainty described Bayesian [3]. answer best context entails answer. Dually, also compute, query, source; is, highest query. Similar [4], PNs graphical models providing representation discrete distribution, through some independence assumptions [2]. A Ω function Pos : → [0, 1] intuitively provides how possible ω ∈ happen. This extended sets Γ ⊆ defining Pos(Γ ) = supω∈Γ Pos(ω). product defined equation ∩Θ) | Θ) · Pos(Θ). decompose probability distributions depend structure graph.