Labeled Bipolar Argumentation Frameworks
Authors: Melisa G. Escañuela Gonzalez, Maximiliano C. D. Budán, Gerardo I. Simari, Guillermo R. Simari
JAIR 2021 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, a polynomial-time algorithm to perform the labeling process is introduced, in which the argument interactions are considered. ... Furthermore, we present the algorithms to perform the labeling process. ... In this work, based on the previous analysis, we have extended BAF by taking into account the properties associated with the arguments in the form of labels, increasing the representational capabilities of this formalization. These labels can be combined and propagated through the bipolar argumentation graph in accordance with the arguments interaction. Then, considering the additional information provided by these labels, the semantics offered by BAF is improved by: (i) obtaining more information about the arguments, (ii) defining new acceptability-based extensions, and (iii) establishing user postures to analyze the argumentation framework where special constraints are satisfied, In previous works by Budán et al. (2014, 2016), we presented preliminary results related to this line of research; here, we improve previous semantic definitions, present an algorithm to perform the labeling process, and analyze the set of postulates that satisfy the semantics presented in this formalism. ... Analyzing L-BAF Ψ (cf. Figure 6) using the classical bipolar notions, we have the extensions obtained in the Example 1, where S1 = {I,K,E,C,H} is d-admissible, s-admissible, and c-admissible, while S2 = {I,K,E, C,D,J,H} is d-admissible, but not s-admissible and not c-admissible. Recall that the set S2 is a d-preferred extension, since it is the maximal d-admissible set; however, in this case, there exists neither non-trivial s-preferred extension, nor non-trivial c-preferred extension. |
| Researcher Affiliation | Academia | Melisa G. Esca nuela Gonzalez EMAIL Departamento de Materm atica Universidad Nacional de Santiago del Estero (UNSE) Consejo Nacional de Investigaciones Cient ıficas y T ecnicas (CONICET) Sgo. del Estero, Argentina Maximiliano C. D. Bud an EMAIL Departamento de Materm atica Universidad Nacional de Santiago del Estero (UNSE) Consejo Nacional de Investigaciones Cient ıficas y T ecnicas (CONICET) Sgo. del Estero, Argentina Gerardo I. Simari EMAIL Departamento de Ciencias e Ingenier ıa de la Computaci on Universidad Nacional del Sur (UNS) Instituto de Ciencias e Ingenier ıa de la Computaci on (ICIC UNS-CONICET) Bah ıa Blanca, Argentina Guillermo R. Simari EMAIL Departamento de Ciencias e Ingenier ıa de la Computaci on Universidad Nacional del Sur (UNS) Instituto de Ciencias e Ingenier ıa de la Computaci on (ICIC UNS-CONICET) Bah ıa Blanca, Argentina |
| Pseudocode | Yes | Algorithm 1: Labeling procedure for a bipolar graph GΘ Algorithm 2: Labeling function for an argument in the bipolar graph |
| Open Source Code | No | Current and future work involves developing an implementation of L-BAF by instantiating it in the existing De LP (Garc ıa & Simari, 2014) system as a basis; the resulting implementation will be evaluated in different domains that require extra information associated with arguments, taking as motivation studies and analyses such as P-De LP (Alsinet et al., 2008a, 2008b). |
| Open Datasets | No | We now introduce a running example showing how our approach could help to achieve an improved analysis, as discussed in the previous section. Consider a scenario involving a discussion by a group of parents on Twitter (or a similar social platform) analyzing the adequacy of a school for a child. In order to reach a decision, they evaluate arguments according to: (i) their preferences, assigning to each of them an assessment of relevance, and (ii) a degree of social rating associated with each argument, describing how popular the argument is in the discussion. Arguments A through K are obtained from the tweets depicted in Figure 1; we assume that the social rating can be derived from the tweet s performance (retweets, likes, and comments), which is shown on the right. ... It is important to remark that the process of identifying relevant tweets, and obtaining arguments from them, is out of the scope of this paper |
| Dataset Splits | No | The paper uses a running example with hypothetical arguments and their initial valuations (Table 3), derived from tweets shown in Figure 1. There is no mention of dataset splits like training, validation, or test sets, as the work demonstrates a theoretical framework and its application to a single illustrative scenario. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running experiments. It focuses on the theoretical framework and algorithmic procedures demonstrated with a running example. |
| Software Dependencies | No | The paper mentions that future work involves instantiating L-BAF in the existing De LP system, but it does not specify any software dependencies with version numbers for the work presented in this paper. |
| Experiment Setup | No | The paper defines an algebra of argumentation labels with specific operators (Definition 8) and provides initial argument valuations for a running example (Table 3), but these are fundamental components of the framework itself rather than experimental setup details like hyperparameters or training configurations for a computational model. |