Alessandro Petruzzelli
A PhD student in computer science.
Currently
1 Oct 2023 - Present
P.hD Student in Computer Science
University of Bari “Aldo Moro”, Bari, IT
My PhD research focuses on the intersection of large language models (LLMs), recommender systems, and conversational recommender systems. I explore how LLMs can be leveraged to enhance recommendation accuracy and personalize user experiences through natural language conversations.
Research interests
Recommender System, Conversational Recommender, Representation Techniques, Large Language Model
Education
2013-2018
Economic Technical Institute High School Diploma
Istituto tecnico Economico “E. Carafa”, Andria, IT
Finale grade: 94/100
24 Sep 2018 - 15 Jul 2021
Bachelor Degree in Computer Science
University of Bari “Aldo Moro”, Bari, IT
Thesis: Design and Development of a Content-based Recommender System to Support a Conversational Agent
Finale grade: 110/110 with honors
27 Sep 2021 - 19 Jul 2023
Master Degree in Data Science
University of Bari “Aldo Moro”, Bari, IT
Thesis: Transformer-Based Conversational Recommendation based on Knowledge Graph
Finale grade: 110/110 with honors
Projects
Design and Development of a Content-based Recommender System to Support a Conversational Agent
With the aim of providing an alternative to a graph-based recommendation system, the project focused on the implementation and comparison of several models for content representation. Thanks to the implementation of these models, it was possible to implement a content-based recommendation system that, in addition to being more effective, allows the implementation of a process of explaining the recommendation. In particular, for the project, vector representation models of textual content were evaluated, thus providing a preprocessing through Natural Language Processing techniques.
Transformer-Based Conversational Recommendation based on Knowledge Graph
Conversational Recommender Systems (CRSs) have attracted a lot of attention in recent years for their ability to provide personalized recommendations through multi-turn interactions. This study is based on an in-depth analysis of the state-of-the-art solutions, focusing on two main areas: recommendation systems that use deep learning techniques and end-to-end conversational systems. The project described in this study falls into the second category. The proposed idea is to use a Transformer architecture to develop a CRS. By exploiting self-attention, this solution is able to learn a representation for each element to be recommended. However, the experiments conducted for this study have shown that the integration of external knowledge from a Knowledge Graph can improve the accuracy of the model if it is integrated into the early encoding layers of the architecture.
25 Feb 2023 - 30 Sep 2023
Scholarship “ANALYSIS AND VALIDATION OF RECOMMENDATION METHODS AND ALGORITHMS FOR THE PERSONALIZATION OF SUGGESTIONS IN THE RETAIL SECTOR”
University of Bari “Aldo Moro”, Bari, IT
Through a careful analysis of purchase data, consumer preferences, and behavioral patterns, the challenges and opportunities for improving the personalization of recommendations were identified. During the research process, deep learning and data analysis methodologies were applied to refine existing recommendation algorithms and develop new solutions. Thanks to these methodologies, more accurate and relevant recommendation models were created, able to provide personalized suggestions to consumers based on their specific interests and preferences.