Educational objectives Understand the basic concepts of knowledge representation in Artificial Intelligence. Understand the reference languages for logic-based knowledge representation, particularly the families of class-based and rule-based knowledge representation languages, and the main reasoning techniques for these languages. Understand the main technologies based on these knowledge representation formalisms, particularly RDF and OWL, with the aim of designing and managing an ontological knowledge base. Understand approaches to numerical representation (embedding) of logical-symbolic knowledge (knowledge graphs) and their processing. Understand the elements of knowledge representation and reasoning in non-symbolic Artificial Intelligence applications, particularly in generative AI and Large Language Models.
Knowledge and Understanding
• Recall and define the fundamental concepts and principles that guide knowledge representation (KR) within computer systems.
• Understand the syntax and formal semantics of class-based and rule-based logic formalisms.
• Understand the expressive limitations of ALC, the open-world assumption typical of Description Logics, and the structural differences from First-Order Logic (FOL) and SQL.
• Understand the operation of the fundamental automated reasoning algorithms for Description Logics and Datalogs.
• Recall and understand the Semantic Web standards, technologies, and protocols defined by the W3C, including RDF, RDF Schema (RDFS), SPARQL, OWL, and the various profiles of OWL 2.
• Understand the structure and properties of Knowledge Graphs, the machine learning methodologies applied to them, and the concept of Knowledge Graph embedding.
• Understand the intersection between symbolic knowledge representation and Large Language Models (LLM), analyzing the latter's actual logical reasoning capabilities and the integration of OWL and Datalog within them.
Applying knowledge and understanding
• Apply logical formalisms (ALC, Datalog, and ASP) to model real-world knowledge domains, translating application requirements into formal constraints, rules, and axioms.
• Apply the Tableau algorithm to manually or assistively solve logical reasoning tasks on TBoxes and ABoxes expressed in ALC.
• Apply and implement Datalog and ASP programs to solve reasoning problems.
• Apply the RDF, RDFS, and OWL standards for the structured design of ontologies and knowledge graphs, formulating complex semantic queries using the SPARQL language.
• Apply the technical specifications of the OWL 2 profiles (OWL 2 QL, OWL 2 RL, and OWL 2 EL), making appropriate use of query rewriting or ABox materialization to optimize querying and data access based on the selected profile.
• Effectively use (and apply) professional and widely used software tools, such as the Protégé ontology editor, for modeling and validating real-world ontologies.
• Create and develop a practical and autonomous knowledge engineering project, integrating semantic technologies, OWL ontologies, logical rule engines, or knowledge graphs to solve a specific application problem.
Making judgments
• Critically analyze and compare the computational, expressive, and structural differences between different data and knowledge management models (SQL vs. Description Logic vs. Datalog vs. ASP).
• Evaluate the computational complexity of reasoning operations within Datalog, ASP, and the various OWL 2 profiles, in order to select the most efficient and scalable technological tool and formalism for a given application scenario.
• Critically evaluate the limitations and advantages of systems based on formal logical reasoning compared to modern sub-symbolic approaches based on machine learning, Knowledge Graph Embeddings, and Large Language Models (LLM), understanding when it is preferable to adopt hybrid or purely symbolic solutions.
Communication skills
• Analyze and clearly structure the documentation and presentation of a practical knowledge engineering project, illustrating with formal rigor and appropriate terminology the modeling choices made, the constraints introduced, and the reasoning engines employed.
• Understand how to present and describe complex theoretical concepts (such as the details of a tableau algorithm, query rewriting logic, or Answer Set semantics) to an audience of specialists (computer science engineers) and non-specialists.
Learning skills
• Independently understand and assimilate the evolution of W3C standards and recommendations related to the Semantic Web and cutting-edge semantic technologies.
• Independently analyze the most recent scientific literature and new software tools for knowledge representation, in particular those focused on the synergistic integration between symbolic AI (ontologies, logics) and generative/sub-symbolic AI (LLM and graph machine learning).
|
Educational objectives Given for granted some basic and indispensable goals commonly shared by the Master in Philosophy and in Philosophy and AI, the course intends to attain the following specific objectives: Knowledge and ability to understand (Dublin descriptor A): detailed and articulated knowledge of a philosophical problem in historical perspective; in-depth knowledge of a philosophical period and context with a focus on the long-term nature of the issues addressed; knowledge of key concepts and terms, even very technical ones, in historical perspective. Application skills (descriptor B): ability to understand and to interpret classic texts of the discipline, also with evaluation of the problems of translation; ability of deep historical analysis and of mature and personal theoretical criticism of texts. Autonomy of judgement (descriptor C): ability to reconstruct in detail a historical-philosophical context; ability to argue and gain a personal perspective on the subject in question (also through active participation in seminars). Communication skills (descriptor D): ability to use a technical vocabulary, even a very specifical one; ability to argue with property of language on the treated topics. Learning ability (descriptor E): ability to deepen the issues also in a very personal, autonomous and original way (through personal bibliographic research, specific insights etc...).
|
Educational objectives General Objectives
The goal of the course is to provide an overview of state-of-the-art natural language processing techniques and their applications.
Specific Objectives
Students will learn the principles of automatic language processing, understanding how machines can interpret, generate and respond to human language. This includes topics such as word representation, word and sense embeddings, neural architectures for NLP, machine translation, and more general text generation.
Knowledge and Understanding
- Knowledge of neural network architectures, such as recurrent neural networks and Transformers, used for natural language processing.
- Knowledge of supervised and unsupervised learning methods in NLP.
- Knowledge of lexical and phrasal computational semantics techniques.
- Understanding of language models for interpreting and generating text.
Applying knowledge and understanding:
- How to develop models for understanding language
- How to develop models for generating language
- How to use neural architectures for NLP
Autonomy of Judgment
Students will be able to evaluate the effectiveness of NLP techniques in different applications.
Communication Skills
Students will be able to explain the principles and techniques of natural language processing.
Next Study Abilities
Students interested in research will discover what are the main open challenges in the area of NLP, obtaining the necessary foundation for more in-depth studies in the field.
|
Educational objectives By the end of the course, students will have achieved the following learning outcomes.
Knowledge and understanding:
knowledge of the theoretical and mathematical principles underlying the main generative models (variational autoencoders, GANs, normalizing flows, diffusion models, large language models);
understanding of the deep learning architectures and the attention mechanism at the core of modern generative models;
knowledge of generation techniques for images, text and multimodal data.
Applying knowledge and understanding:
being able to implement and train generative models across different domains (images, text);
being able to apply alignment and adaptation techniques for language models (RLHF, DPO, LoRA) and Retrieval-Augmented Generation;
being able to use advanced Python libraries (e.g., PyTorch) to design generative systems.
Making judgements:
being able to critically evaluate the performance of generative models using appropriate metrics;
being able to analyse the limitations, risks and potential of generative models in different application contexts.
Communication skills:
being able to describe and justify design choices and experimental results using appropriate technical terminology.
Learning skills:
being able to independently update one's knowledge by following the scientific literature of a rapidly evolving field.
|
Educational objectives Allow the student to support training activity conforms to their individual research interests.
|
Educational objectives The certification of linguistic competence required (Idoneità) is linked to the thesis work. To get the certification, you will need to submit a written paper (in Italian) – between 4.000 and 5.000 characters long – based on a text (article, book, essay) relating to the subject of the thesis work and chosen in agreement with your supervisor. The supervisor will read and evaluate your paper and, in case, will deliver a certificate of linguistic competence. In order to register the exam, you will need to submit the certificate and your Infostud test reservation to one of the members of the Commission.
The Commission is composed of: prof. Marco Mazzeo.
|
Educational objectives The final test involves the preparation, presentation and discussion before a committee of a paper in Italian or in English, prepared independently by the student. At the final examination are awarded 24 credits.
|
