Argumentation is a major topic in the study of artificial intelligence. In particular, the problem of solving certain reasoning tasks on Dung's abstract argumentation frameworks is central to many advanced argumentation systems. The fact that problems to be solved are intractable requires efficient algorithms and solvers.
The ICCMA competition aims at nurturing research and development of implementations for computational models of Argumentation. The objectives of the competition are to provide a forum for empirical comparison of solvers, to highlight challenges to the community, to propose new directions for research and to provide a core of common benchmark instances and a representation formalism that can aid in the comparison and evaluation of solvers.
The Third International Competition on Computational Models of Argumentation (ICCMA'19) will focus on reasoning tasks in abstract argumentation frameworks. Submitted solvers will be tested on a selected collection of benchmark instances.
Solvers will be evaluated based on their performance in solving the following problems on Abstract Argumentation Frameworks (AFs):
- (SE) Given an AFs, determine some extension
- (EE) Given an AFs, determine all extensions
- (DC) Given an AFs and some argument, decide whether the given argument is credulously inferred
- (DS) Given an AFs and some argument, decide whether the given argument is skeptically inferred
The above computational problems are to be solved with respect to the following standard semantics:
- (CO) Complete Semantics
- (PR) Preferred Semantics
- (ST) Stable Semantics
- (SST) Semi-stable Semantics
- (STG) Stage Semantics
- (GR) Grounded Semantics (only SE and DC)
- (ID) Ideal Semantics (only SE and DC)
In the competition, there will be a track for each of these 7 semantics. In addition, a special track will be conducted.
This track is focused on solvers processing dynamic frameworks, i.e., solvers described in terms of changes w.r.t. previous ones: a solver can reuse the solution obtained previously to be faster on the same framework modulo a new argument or attack. We will provide two different types of input for the changes.
Option 1. AFs will be passed to solvers together with a list of changes in a separate file. Changes may consist of a sequence of random additions/deletions of attacks, provided through a simple text format, e.g.,
+att(a,b). -att(d,e). ..., representing the introduction and deletion of an attack, respectively (two AFs to process in the end).
Option 2. An initial AF, together with a list of modified AFs will be passed as input to the solver. Each modified AFs will be different from the previous in the list for only one single attack.
We will then expect as many outputs as the number of changes plus one (the original framework). The track will consider the four semantics CO, PR, ST, and GR.
ICCMA'19 reasoning tasks are detailed in the Call for Solvers.
Challenging and representative benchmarks are essential to perform significant comparisons of solvers. We invite submissions of both real world benchmarks and benchmark generators to ensure a diverse benchmark set for the competition. In the case of (randomly) generated benchmarks we invite to submit the generator instead of the instances.
All submitted benchmark instances, together with those of previous editions of ICCMA, will become part of the suite of benchmark instances made available to the community after the event, and a selection will be used to evaluate solvers at ICCMA'19.
All the details on Call for Benchmarks.