Attention is an increasingly popular mechanism used in a wide range of neural architectures. The itself has been realized variety formats. However, because the fast-paced advances this domain, systematic overview attention still missing. In article, we define unified model for architectures natural language processing, with focus on those designed to work vector representations textual data. We propose taxonomy models according four dimensions: representation input, compatibility function,...

Argumentation mining aims at automatically extracting structured arguments from unstructured textual documents. It has recently become a hot topic also due to its potential in processing information originating the Web, and particular social media, innovative ways. Recent advances machine learning methods promise enable breakthrough applications economic sciences, policy making, technology: something that only few years ago was unthinkable. In this survey article, we introduce argumentation...

SCIFF is a framework thought to specify and verify interaction in open agent societies. The language equipped with semantics based on abductive logic programming; SCIFF's operational component new programming proof procedure, also named SCIFF, for reasoning expectations dynamic environments. In this article we present the declarative of language, termination, soundness, completeness results demonstrate possible application multiagent domain.

The automatic extraction of arguments from text, also known as argument mining, has recently become a hot topic in artificial intelligence. Current research only focused on linguistic analysis. However, many domains where communication may be vocal or visual, paralinguistic features too contribute to the transmission message that intend convey. For example, political debates crucial role is played by speech. question we address this work whether such one can improve claim detection for...

This report contains preliminary results of the study aiming at automating legal evaluation privacy policies, under GDPR, using artificial intelligence (machine learning), in order to empower civil society representing interests consumers. We outline what requirements a GDPR-complaint policy should meet (comprehensive information, clear language, fair processing), as well are ways which these documents can be unlawful (if required information is insufficient, language unclear, or potentially...

Current research in machine learning and artificial intelligence is largely centered on modeling performance evaluation, less so data collection. However, recent demonstrated that limitations biases may negatively impact trustworthiness reliability. These aspects are particularly impactful sensitive domains such as mental health neurological disorders, where speech used to develop AI applications for patients healthcare providers. In this paper, we chart the landscape of available datasets...

The concept of an agent is increasingly used in contemporary software applications, particularly those involving the Internet, autonomous systems, or cooperation. However, with dependability and safety mind, it vital that mechanisms for representing implementing agents are clear consistent. Hence there has been a strong research effort directed at using formal logic as basis descriptions implementation. Such logical not only presents clarity consistency required but also allows important...

In open societies of agents, where agents are autonomous and heterogeneous, it is not realistic to assume that will always act so as comply with interaction protocols. Thus, the need arises for a formalism specify constraints on agent interaction, tool able observe check compliance this paper we present JAVA-PROLOG software component built logic programming technology, which can be used verify protocols, has been integrated PROSOCS platform.

In multiagent systems, agent interaction is ruled by means of protocols. Compliance to protocols can be hardwired in programs; however, this requires that only "certified" agents interact. open societies, composed autonomous and heterogeneous whose internal structure is, general, not accessible, should specified terms the observable behaviour, compliance verified an external entity.In paper, we propose a Java-Prolog-CHR system for verification agents' behaviour logic-based formalism (Social...

Abstract Recent work has demonstrated how data-driven AI methods can leverage consumer protection by supporting the automated analysis of legal documents. However, a shortcoming approaches is poor explainability. We posit that in this domain useful explanations classifier outcomes be provided resorting to rationales. thus consider several configurations memory-augmented neural networks where rationales are given special role modeling context knowledge. Our results show not only contribute...

Since its introduction, the Event Calculus (ℰ𝒞) has been recognized for being an excellent framework to reason about time and events, it applied a variety of domains. However, formalization inside logic-based frameworks

We explore the use of residual networks for argumentation mining, with an emphasis on link prediction. The method we propose makes no assumptions document or argument structure. evaluate it a challenging dataset consisting user-generated comments collected from online platform. Results show that our model outperforms equivalent deep network and offers results comparable state-of-the-art methods rely domain knowledge.

We explore the use of residual networks and neural attention for multiple argument mining tasks. propose a architecture that exploits attention, multi-task learning, makes ensemble, without any assumption on document or structure. present an extensive experimental evaluation five different corpora user-generated comments, scientific publications, persuasive essays. Our results show our approach is strong competitor against state-of-the-art architectures with higher computational footprint...