Semantics in Intelligent Systems

Course pre-requisite(s): Basics of Mathematics, Probability, Statistics. Knowledge of Programming Languages.

Course Overview
The goal of the course is to provide students with an overview of the different state-ofthe-art strategies to handle structured and unstructured textual content, and to encode semantics-aware content representation in intelligent systems, such as user profiling
platforms, recommender systems, information retrieval and question answering systems. The course will start with basic notions about NLP and content representation, then it will follow with a discussion of more advanced and sophisticated methodologies, including recent knowledge graph embeddings and deep learning-based transformers, and will show how these techniques are applied to intelligent systems to improve predictive accuracy and overall user experience.

Learning Outcomes
By the end of this course, the learner will acquire knowledge about the fundamentals of semantics and intelligent systems. In particular, the learner will be able to know and understand:
- the theoretical, methodological, and operational aspects of semantics with particular reference to the main levels of content representation;
- the techniques and the main open-source frameworks for semantics and intelligent systems;
- the theoretical, methodological, and operational aspects of intelligent systems and semantic information indexing and sharing;
- the theoretical, methodological, and operational aspects of information filtering systems and recommender systems;
- the techniques and major open-source frameworks for the design of recommender systems.

Course Content
1. Semantics, NLP and content representation: basics. (Lecture: 4 hours)
2. Semantics, NLP and content representation: advanced techniques. (Lecture: 4 hours - Hands-on: 2 hours.)
3. Encoding endogenous semantics. (Lecture: 3 hours)
4. Encoding exogenous semantics (Lecture: 3 hours)
5. Intelligent systems based on semantics: semantics-aware recommender systems. (Lecture: 2 hours - Hands-on: 2 hours.)
6. Intelligent systems based on semantics: semantics-aware social media analysis. (Lecture: 2 hours - Hands-on: 2 hours.)
7. Intelligent systems based on semantics: advanced topics and challenges (deep learning, explanations, conversational systems, serendipity). (Lecture: 4 hours - Hands-on: 2 hours)
8. Individual/Group Project – 12 hours

Instructional Method
Most of the course will rely on lectures and seminars on basic and advanced topics related to content representation (22 hour). Then, 8 hours will be used as hands-on, to allow students to apply the concepts to concrete use-cases. Finally, 12 hours will be used to allow students to select and develop their own group projects, whose output will be used as final assessment of the course.

Required Course Materials
All the required materials will be shared throughout the course as PDF or Powerpoint files. Moreover, students are encouraged to read the book “Semantics in Adaptive and Personalized Systems: Methods, Tools and Applications'', edited by Springer”, since
most of the materials rely on the book.
As for the hands-on part, Python is required. Notebooks will be shared for the exercises.

Assessment
Each project work developed by a student/group of students will be evaluated and a grade will be assigned at the end of the course.