AIxIA 2021 – Advances in Artificial Intelligence: 20th International Conference of the Italian Association for Artificial Intelligence, Virtual Event, ... (Lecture Notes in Computer Science, 13196)

Preface
Organization
Contents
Planning and Strategies
Task Allocation for Multi-robot Task and Motion Planning: A Case for Object Picking in Cluttered Workspaces
	1 Introduction
	2 Background
		2.1 Related Work
		2.2 Task-Motion Planning and AND/OR Graphs
		2.3 Problem Definition
	3 Multi-robot Task and Motion Planning
		3.1 Obstacles Selection
		3.2 Task Allocation
		3.3 Task Decomposition
		3.4 Task and Motion Planning
		3.5 The Multi-robot Task-Motion Planning Loop
	4 Experimental Results
	5 Conclusion
	References
A Sound (But Incomplete) Polynomial Translation from Discretised PDDL+ to Numeric Planning
	1 Introduction
	2 Background
		2.1 Polynomial Translation
	3 POLY-: A Sound (but Incomplete) Translation
	4 Experimental Analysis
	5 Conclusion
	References
Enhancing Telepresence Robots with AI: Combining Services to Personalize and React
	1 Introduction
	2 An Architecture to Enhance Robotic Telepresence
	3 The Deliberative Services for Personalization
		3.1 Executing Plans and Adapting Them to the Reality
	4 Contextualized Navigation Services
	5 Advanced Perception
		5.1 Emotion Detection
	6 The Integrated System at Work
	7 Conclusions
	References
Tafl-ES: Exploring Evolution Strategies for Asymmetrical Board Games
	1 Introduction
	2 Background
		2.1 Hnefatafl
		2.2 State of the Art
		2.3 Evolution Strategies
	3 Approach
		3.1 Problem Formulation
		3.2 Proposed Solution: Tafl-ES
	4 Results
	5 Conclusions and Future Work
	References
Constraints, Argumentation, and Logic Programming
Combining DCOP and MILP for Complex Local Optimization Problems
	1 Introduction
	2 Adopted Techniques and Related Work
		2.1 Distributed Constraint Optimization Problem
		2.2 DCOP with Complex Local Problems
		2.3 Supply Chain Coordination Problem (SCC)
		2.4 Attempts to Solve SCC with DCOP
	3 Combining DCOP with MILP to Solve SCC
		3.1 Formal Model
		3.2 Algorithms
	4 Experiments
		4.1 Description of Experiments
		4.2 Results Analysis
	5 Conclusions and Future Work
	References
Automated Design of Elevator Systems: Experimenting with Constraint-Based Approaches
	1 Introduction
	2 Design of Elevator Systems
	3 Encoding Elevator Systems Design
	4 Experimental Results
	5 Conclusions
	References
Modular Logic Argumentation in Arg-tuProlog
	1 Introduction
	2 The Domain of Private International Law: Running Examples
	3 Modular Argumentation in Arg-tuProlog
		3.1 Modular Logic: Architecture and Predicates
	4 Running Examples in Arg-tuProlog
	5 Conclusion
	References
Burden of Persuasion in Meta-argumentation
	1 Introduction
	2 Meta-argumentation Framework
		2.1 Structured Argumentation for Object-Level Argumentation
		2.2 Object and Meta Level Connection: Bimodal Graphs
		2.3 Argument Schemes for Meta-level Argumentation
	3 Burden of Persuasion as Meta-argumentation
		3.1 Meta-level Graph
		3.2 Object and Meta Level Connection: Supporting Sets
		3.3 Equivalence with Burden of Persuasion Semantics
	4 Burden Inversion
	5 Conclusions
	References
Knowledge Representation, Reasoning, and Learning
Reasoning About Smart Contracts Encoded in LTL
	1 Introduction
	2 Preliminaries
		2.1 Smart Contracts
		2.2 Linear Temporal Logic with Past Operators
	3 Encoding Smart Contracts in Linear Temporal Logic
	4 Reasoning About Smart Contracts
		4.1 Reasoning Problems on a Single Smart Contract
		4.2 Reasoning Problems on a Set of Smart Contracts
	5 SCRea: Smart Contracts Reasoner
		5.1 Conceptual Architecture
		5.2 Encoder and Decoder Modules
		5.3 SCRea at Runtime
	6 Concluding Remarks
	References
A Combinatorial Approach to Weighted Model Counting in the Two-Variable Fragment with Cardinality Constraints
	1 Related Work
	2 FOMC for Universal Formulas
	3 FOMC for Cardinality Constraints
	4 FOMC for Existential Quantifiers
	5 Weighted First-Order Model Counting
	6 Conclusion
	References
Option Discovery for Autonomous Generation of Symbolic Knowledge
	1 Introduction
	2 Problem Description
	3 Implementation
		3.1 Option Discovery
		3.2 Abstracting Options in PPDDL
	4 Empirical Analysis
	5 Conclusions and Future Work
	References
Natural Language Processing
A Neural-Machine-Translation System Resilient to Out of Vocabulary Words for Translating Natural Language to SPARQL
	1 Introduction
	2 Preliminaries
		2.1 Knowledge Bases and SPARQL
		2.2 Recurrent Neural Networks
	3 From Natural Language Questions to SPARQL
		3.1 Mitigating the WOOV Problem
		3.2 Out-of-Vocabulary Words in NL Questions
		3.3 The Model
	4 Experiments
		4.1 Evaluation on Monument Dataset
		4.2 Evaluation on QALD-9
	5 Related Work
	6 Conclusions and Future Work
	References
Exploiting Textual Similarity Techniques in Harmonization of Laws
	1 Introduction
	2 Related Work
	3 Case Study
		3.1 CrossJustice Project
		3.2 Types of Annotations
		3.3 Dataset Overview
	4 Methodology
		4.1 Text Processing
		4.2 Similarity Measure
	5 Output
		5.1 Text Representation
		5.2 Heat Maps Visualization
	6 Conclusions
	References
Easy Semantification of Bioassays
	1 Introduction
	2 A Motivating Example for Bioassay Semantification
	3 Related Work
		3.1 Corpora of Semantified Life Science Publications
		3.2 AI-Based Scholarly Knowledge Graph Construction
	4 Materials and Methods
		4.1 An Expert-Annotated Semantified Bioassays Corpus
		4.2 Labeling Task Definition for Bioassay Semantification
		4.3 Clustering Task Definition for Bioassay Semantification
	5 Bioassay Semantification Experiments
		5.1 Experimental Setup
		5.2 Experimental Results
	6 Digital Library Bioassay Semantification Workflows
	7 Conclusion
	References
Pruned Graph Neural Network for Short Story Ordering
	1 Introduction
	2 Related Work
		2.1 Sentence Ordering
		2.2 Graph Neural Networks in NLP
	3 Baselines
		3.1 ATTOrderNet
		3.2 SE-Graph
	4 Methodology
		4.1 Problem Formulation
		4.2 Dataset
		4.3 Pruned Graph Sentence Ordering (PG)
		4.4 Majority Voting
	5 Experiment
		5.1 Evaluation Metrics
		5.2 Contrast Models
		5.3 Setting
		5.4 Results
	6 Conclusion
	References
Multi-task and Generative Adversarial Learning for Robust and Sustainable Text Classification
	1 Introduction
	2 Multi-task and Generative Adversarial Learning in MT-GAN-BERT
	3 Experimental Evaluation
	4 Conclusion
	References
Punctuation Restoration in Spoken Italian Transcripts with Transformers
	1 Introduction
	2 Related Work
	3 Experimental Setting
		3.1 Model
		3.2 Data
	4 Results
	5 Error Analysis
	6 Extended Evaluation
	7 Conclusions
	References
AI for Content and Social Media Analysis
On the Impact of Social Media Recommendations on Opinion Consensus
	1 Introduction
	2 The Model
	3 Symmetric Two-Block Model
		3.1 Characterization
	4 General Networks
	5 Conclusions
	References
Misogynous MEME Recognition: A Preliminary Study
	1 Introduction
	2 State of the Art
	3 The MEME Dataset
	4 Models and Results
		4.1 Unimodal Classifiers
		4.2 Multimodal Classifier
	5 Conclusions
	References
Signal Processing: Images, Videos and Speech
A Relevance-Based CNN Trimming Method for Low-Resources Embedded Vision
	1 Introduction
	2 Related Work
	3 The Pruning Method
	4 Applications and Experimental Evaluation
		4.1 Pruning CNNs for Camera Tasks
		4.2 VGG16 on Cifar10
	5 Conclusions
	References
Vision-Based Holistic Scene Understanding for Context-Aware Human-Robot Interaction
	1 Introduction
	2 Related Work
	3 Datasets
	4 Methodology
		4.1 Convolutional Neural Network
		4.2 Recurrent Neural Network
		4.3 CNN + RNN Architecture
	5 Experiments
		5.1 Dataset
		5.2 Implementation Details
		5.3 Result Analysis
	6 Context-Aware Human-Robot Interaction
		6.1 Application
	7 Conclusion and Further Work
	References
Human Detection in Drone Images Using YOLO for Search-and-Rescue Operations
	1 Introduction
	2 Related Work
	3 Materials and Methods
		3.1 HERIDAL Dataset
		3.2 SARD Dataset
		3.3 YOLOv5
	4 Experiment
		4.1 Setting
		4.2 Metrics
		4.3 Results
	5 Conclusion
	References
ArabCeleb: Speaker Recognition in Arabic
	1 Introduction
	2 The ArabCeleb Dataset
		2.1 Description
		2.2 Collection Pipeline
	3 Experiments
		3.1 Speaker Verification Baseline Methods Considered
		3.2 Results
	4 Conclusions
	References
Static, Dynamic and Acceleration Features for CNN-Based Speech Emotion Recognition
	1 Introduction and Related Work
	2 Proposed Method
		2.1 Static, Dynamic and Acceleration Features
		2.2 Global Features
		2.3 Proposed 1-D Convolutional Neural Network
	3 Experiments
		3.1 Datasets
		3.2 Comparison with the State of the Art
		3.3 Result
	4 Conclusion
	References
EEG-Based BCIs for Elderly Rehabilitation Enhancement Exploiting Artificial Data
	1 Introduction
	2 Related Works
	3 Methods
		3.1 Datasets
		3.2 Multivariate Empirical Mode Decomposition
	4 Our Proposal
	5 Results and Discussion
	6 Conclusions
	References
Machine Learning for Argumentation, Explanation, and Exploration
Supporting Trustworthy Artificial Intelligence via Bayesian Argumentation
	1 Introduction
	2 A Primer in Statistical Learning
	3 A Primer in Bayesian Argumentation
	4 Argumentative-Generative Framework
	5 Conclusions
	References
Logic Constraints to Feature Importance
	1 Introduction
	2 Bibliographic Review
	3 Mathematical Setting of Feature Importance
	4 Constraints to Feature Importance
	5 Fairness Through Feature Importance Constraints
	6 Toy Example: Constraint of the Form Lg
	7 Fairness Through Constraints to Feature Importance
	8 Conclusion and Future Work
	References
Clustering-Based Interpretation of Deep ReLU Network
	1 Introduction
	2 Bibliographic Review
	3 Deep ReLU Networks for the Partition of the Input Space
	4 Simulation Study
	5 Titanic Dataset
	6 Conclusions and Limitations
	References
Exploration-Intensive Distractors: Two Environment Proposals and a Benchmarking
	1 Introduction
	2 Related Work
		2.1 Approaches to Sparsity
		2.2 Distractors
	3 Environment Design
		3.1 Base Environment
		3.2 Single-Action Static-Rendering (SASR) TV
		3.3 Multi-Action Dynamic-Rendering (MADR) TV
		3.4 ViZDoom-TV Integration
	4 Benchmarking
		4.1 Covered Algorithms
		4.2 Training Details
		4.3 Results and Discussion
	5 Conclusions
	References
Neural QBAFs: Explaining Neural Networks Under LRP-Based Argumentation Frameworks
	1 Introduction
	2 Background
		2.1 MLP Basics
		2.2 LRP Basics
		2.3 QBAF Basics
	3 nQBAFS and LRP-Based Argumentation Semantics
	4 Properties for nQBAFS Under LRP Semantics
	5 Empirical Study
		5.1 DAX Basics
		5.2 The Basics of Google's Method
		5.3 Settings
		5.4 DAX Vs Google Comparisons
		5.5 Discussion
	6 Conclusions
	References
Machine Learning and Applications
Domino Saliency Metrics: Improving Existing Channel Saliency Metrics with Structural Information
	1 Introduction
	2 Data Flow Graph for Pruning
		2.1 Background
		2.2 Channel Pruning Networks with Splits and Joins
		2.3 Data Flow Graph
		2.4 Join and Split Nodes
		2.5 Group Convolution
		2.6 How to Prune Biases and Activation Layers
	3 Domino Pruning
	4 Experimental Evaluation
		4.1 Pruning Algorithm
		4.2 Networks
		4.3 Saliency Metrics
	5 Results
	6 Related Work
	7 Conclusion
	References
Learned Sorted Table Search and Static Indexes in Small Model Space
	1 Introduction
		1.1 Learned Searching in Sorted Sets
		1.2 Our Contributions
	2 A Simple View of Learned Searching in Sorted Sets
	3 Experimental Methodology
		3.1 Sorted Table Search and Classic Indexes
		3.2 Model Classes Characterizing Model Space
		3.3 Hardware
		3.4 Datasets
	4 Learning the CDF of a Sorted Table and Mining SODS Output for the Synoptic RMI: Outline of Experiments and Findings
	5 Constant Space Models: Outline of Query Experiments
	6 Parametric Space Models: Outline of Query Experiments
	7 Conclusions and Future Directions
	References
Siamese Networks with Transfer Learning for Change Detection in Sentinel-2 Images
	1 Introduction
	2 Related Work
	3 The Proposed Approach
		3.1 Siamese Network
		3.2 Fine-tuning
	4 Implementation Details
	5 Experimental Evaluation
		5.1 Datasets, Experimental Setting and Evaluation Metrics
		5.2 Results
	6 Conclusion
	References
Adversarial Machine Learning in e-Health: Attacking a Smart Prescription System
	1 Introduction
	2 Related Work
	3 Case Study
	4 Threat Model
	5 Methodology
	6 Experimental Analysis
		6.1 The Classification Network
		6.2 Results and Discussion
	7 Conclusions
	References
Deep Learning of Recurrence Texture in Physiological Signals
	1 Introduction
	2 Methods
		2.1 Texture of Recurrence Dynamics
		2.2 Deep Learning of Texture of Recurrence in Time Series
		2.3 Simulation of Texture for Data Augmentation in Deep Learning
		2.4 Procedure of Proposed Method
	3 Results and Discussion
	4 Conclusion
	References
Highlighting the Importance of Reducing Research Bias and Carbon Emissions in CNNs
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Base Network Selection
		3.2 Categories for Experimentation
		3.3 How do we Measure Energy Consumption?
	4 Experimental Evaluation
		4.1 Architecture Modification
		4.2 Learning Rate Scheduler
		4.3 Data Augmentation
		4.4 Optimizer
		4.5 Loss Function
		4.6 Custom Nodes and Layers
	5 Discussion
	References
Generating Local Textual Explanations for CNNs: A Semantic Approach Based on Knowledge Graphs
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 The Knowledge Graph Model
		3.2 Generation of Textual Factual and Counterfactual Explanations for Mistakes
	4 Experimental Evaluation
		4.1 Evaluating Link Prediction for Semantic Attributes
		4.2 A Visual Evaluation for Factual and Counterfactual Explanations
	5 Conclusions and Future Work
	References
Detection Accuracy for Evaluating Compositional Explanations of Units
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Network Dissection and Compositional Explanations
		3.2 Detection Accuracy
	4 Experiments
		4.1 Setup
		4.2 Detection Accuracy as Evaluation Metric
		4.3 Detection Accuracy as Optimization Metric
	5 Conclusion and Future Work
	References
Knowledge-Based Neural Pre-training for Intelligent Document Management
	1 Introduction
	2 Transformers for Robust NL Inferences
	3 ABILaBERT: Injecting Domain Knowledge in BERT
		3.1 Injecting Process Knowledge as Auxiliary Tasks
		3.2 Auxiliary Tasks for Domain Specific Pre-training
	4 Using ABILaBERT for Text-Driven Process Mining
	5 Experimental Evaluation
	6 Conclusion
	References
Improving Machine Translation of Arabic Dialects Through Multi-task Learning
	1 Introduction
	2 Related Work
	3 Dataset
	4 Methodology
		4.1 Sequence-to-Sequence Learning
		4.2 Multi-task Sequence-to-Sequence Learning
		4.3 Model Training
	5 Experimental Results and Discussion
		5.1 Experiment Settings
	6 Conclusion
	References
Continuous Defect Prediction in CI/CD Pipelines: A Machine Learning-Based Framework
	1 Overview
	2 State of the Art
	3 The Framework
		3.1 The Dataset
		3.2 Data Preparation
		3.3 Feature Engineering and Selection
		3.4 The Models
		3.5 The Monitoring Dashboard
	4 Infrastructural Considerations
	5 Conclusions
	References
AI Applications
Robust Optimization Models For Local Flexibility Characterization of Virtual Power Plants
	1 Introduction
	2 Related Work
	3 VPP: A Distributed Architecture
	4 Model Description
		4.1 Modeling of Uncertainties
		4.2 Model and Components
		4.3 Modeling of RES Production
		4.4 Modeling of Storage Systems
		4.5 External Grid
		4.6 Modeling of Generator Units
		4.7 Demand Side Management
		4.8 Power Balance
		4.9 Objective Functions
	5 Case Studies
		5.1 Dataset Description and Local VPP Configurations
	6 Experimental Results and Discussion
		6.1 Cost Comparison
	7 Conclusion
	References
Explainable Artificial Intelligence for Technology Policy Making Using Attribution Networks
	1 Introduction and Motivation
		1.1 AI-Driven Law and Policy
		1.2 Causality Through Attributions
	2 Methods
		2.1 Deep Learning for Legal Analysis
		2.2 NLP for Law
		2.3 Attribution Networks for Policy
		2.4 Measuring Similarity Using Nearest Neighbors
	3 Experimental Work
	4 Results and Policy Discussions
	5 Conclusion
	References
A Comparative Study of AI Search Methods for Personalised Cancer Therapy Synthesis in COPASI
	1 Introduction
		1.1 Motivation
		1.2 Contribution
		1.3 Paper Outline
	2 State of the Art
	3 VPH Model of the Immune Response to Colorectal Cancer
	4 Modelling the Therapies
	5 Optimisation
		5.1 Optimisation Algorithms
	6 Results
	7 Conclusions
	References
Effective Analysis of Industry-Relevant Cyber-Physical Systems via Statistical Model Checking
	1 Introduction
	2 Industrial Case Studies for Statistical Model Checking
		2.1 Peak Shaving in Smart Grids
		2.2 Virtual Patients for In-Silico Clinical Trials
		2.3 Autonomous Drone Navigation
		2.4 Bluetooth Protocol
		2.5 Leader Election Protocol in IEEE 1394
	3 Conclusions
	References
An ASP-Based Approach to Scheduling Pre-operative Assessment Clinic
	1 Introduction
	2 Problem Description
	3 Formalization of the PAC Scheduling Problem
	4 ASP Encoding
		4.1 ASP Encoding for the First PAC Sub-problem
		4.2 ASP Encoding for the Second PAC Sub-problem
	5 Experimental Results
	6 Domain Specific Optimizations
	7 Related Work
	8 Conclusion
	References
Solving the Dial-a-Ride Problem Using an Adapted Genetic Algorithm
	1 Introduction
	2 Dial-a-Ride Problem
	3 Genetic Algorithm for DARP
	4 Experimental Study
		4.1 Benchmark Setup
		4.2 Results
	5 Conclusion
	References
Unstructured Data in Predictive Process Monitoring: Lexicographic and Semantic Mapping to ICD-9-CM Codes for the Home Hospitalization Service
	1 Introduction
	2 Background
	3 The Home Hospitalization Service Scenario
	4 Approach
		4.1 Data Preprocessing and Analysis
		4.2 Mapping the Diagnosis Field to the ICD-9-CM Dictionary
		4.3 Predicting The Home Hospitalization Outcome
	5 Evaluation
		5.1 ICD-9-CM Mapping Evaluation
		5.2 Home Hospitalization Outcome Prediction Evaluation
	6 Related Work
	7 Conclusions
	References
Author Index