Diapositiva 1

1
Abduction with hypotheses confirmation
Marco Alberti, Marco Gavanelli, Evelina
Lamma Università di Ferrara, Italy malberti,mgava
nelli,elamma_at_ing.unife.it
Paola Mello, Paolo Torroni Università di Bologna,
Italy pmello,ptorroni_at_deis.unibo.it
Abductive Logic Programming

• Example
• Expectations can express various real-life
  concepts
• Expectations on patients conditions
• enquiries, questions to the user, proposal of
  further examinations
• possible solutions, warnings if not adopted
• Declarative semantics
• Abductive semantics
• Consistency
• Fulfilment

Classical ALP extends logic programming with
predicates, called abducibles, that are not
defined, but can be assumed as true. Given a
knowledge base KB and a goal G representing a set
of observations, we want to find a set ? of
possible causes that explains G
Integrity constraints (IC) limit the set of
hypotheses that can be assumed true. Example
Abducibles disease KB goal Answer

• SCIFF Framework
• Extends classical ALP
• Dynamic happening of events
• Generates expected evolution of observed system
• If observations do not match the expected
  behaviour, discard hypotheses
• Universally quantified variables in abducibles
• CLP constraints on existentially and universally
  quantified variables

Abducibles
Resolvent
happened events
Fulfilled expectations
Partially-solved IC
Violated expectations
Pending expectations
Constraint Store
Events Event happened at time T
Expectations
Event expected to happen
Event expected not to happen