Andrea Pazienza

On the Gradual Acceptability of Arguments in Bipolar Weighted Argumentation Frameworks with Degrees of Trust

Computational models of argument aim at engaging argumentation-related activities with human users. In the present work we propose a new generalized version of abstract argument system, called Trust-affected Bipolar Weighted Argumentation Framework (T-BWAF). In this framework, two mainly interacting components are exploited to reason about the acceptability of arguments. The former is the BWAF, which combines and extends the theoretical models and properties of bipolar and weighted Argumentation Frameworks. The latter is the Trust Users Graph, which allow us to quantify gradual pieces of information regarding the source (who is the origin) of an argument. The synergy between them allow us to consider further gradual information which lead to a definition of intrinsic strength of an argument. For this reason, the evaluation of arguments for T-BWAF is defined under a ranking-based semantics, i.e. by assigning a numerical acceptability degree to each argument.

A Similarity-Based Abstract Argumentation Approach to Extractive Text Summarization

Sentence-based extractive summarization aims at automatically generating shorter versions of texts by extracting from them the minimal set of sentences that are necessary and sufficient to cover their content. Providing effective solutions to this task would allow the users to save time in selecting the most appropriate documents to read for satisfying their information needs or for supporting their decision-making tasks. This paper proposes 2 contributions: (i) it defines a novel approach, based on abstract argumentation, to select the sentences in a text that are to be included in the summary; (ii) it proposes a new strategy for similarity assessment among sentences, adopting a different similarity measure than those traditionally exploited in the literature. The effectiveness of the proposed approach was confirmed by experimental results obtained on the English subset of the benchmark MultiLing2015 dataset.

Empowered Negative Specialization in Inductive Logic Programming

In symbolic Machine Learning, the incremental setting allows to refine/revise the available model when new evidence proves it is inadequate, instead of learning a new model from scratch. In particular, specialization operators allow to revise the model when it covers a negative example. While specialization can be obtained by introducing negated preconditions in concept definitions, the state-of-the-art in Inductive Logic Programming provides only for specialization operators that can negate single literals. This simplification makes the operator unable to find a solution in some interesting real-world cases.

An Abstract Argumentation-Based Approach to Automatic Extractive Text Summarization

Sentence-based extractive summarization aims at automatically generating shorter versions of texts by extracting from them the minimal set of sentences that are necessary and sufficient to cover their content. Providing effective solutions to this task would allow the users of Digital Libraries to save time in selecting documents that may be appropriate for satisfying their information needs or for supporting their decision-making tasks. This paper proposes an approach, based on abstract argumentation, to select the sentences in a text that are to be included in its summary. The proposed approach obtained interesting experimental results on the English subset of the benchmark MultiLing 2015 dataset.

Rule Generalization Strategies in Incremental Learning of Disjunctive Concepts

Symbolic Machine Learning systems and applications, especially when applied to real-world domains, must face the problem of concepts that cannot be captured by a single definition, but require several alternate definitions, each of which covers part of the full concept extension. This problem is particularly relevant for incremental systems, where progressive covering approaches are not applicable, and the learning and refinement of the various definitions is interleaved during the learning phase. In these systems, not only the learned model depends on the order in which the examples are provided, but it also depends on the choice of the specific definition to be refined. This paper proposes different strategies for determining the order in which the alternate definitions of a concept should be considered in a generalization step, and evaluates their performance on a real-world domain dataset.

An agent architecture for adaptive supervision and control of smart environments

This paper describes the architecture and functionality of a generic agent that is in charge of handling a given environment in an Ambient Intelligence context, ensuring suitable contextualized and personalized support to the users actions, adaptivity to the users peculiarities and to changes over time, and automated management of the environment itself. The architecture is implemented in a multi-agent system, where different types of agents are endowed with different levels of reasoning and learning capabilities. In addition to controlling normal operations of the environment, the system may identify users needs and goals and activate suitable workflows to satisfy them. Some actions in these workflow involve the execution of semantic services. When a single service is not available for fulfilling a given need, an automatic service composer is used to obtain a suitable combination of services. The architecture has been implemented in a prototypical agent-based system that works in a Smart Home Environment.