Coding projects.

Here a description for every project, from the most to the least recent.

This project for Natural Language Processing (Bocconi) focuses on automating the division of YouTube videos into chapters based on video transcripts. For this task there isn’t a predominant approach in literature, most papers use either topic models or a combination of computer vision and NLP to achieve this goal. Our system aimed at automating chapter creation to enhance user engagement and convenience. We implemented a method that uses LLMs to analyze video transcripts, segment them into chapters, and generate titles, comparing the results with manually created chapters. I worked in collaboration with Emanuele Sala, Luca Soleri and Fabio Stefana.

This project for Simulation and Modeling (Bocconi) looks into the impact of regional economic integration, specifically within the European Union (EU), on global trade networks, focusing on two major events: the introduction of the euro and the United Kingdom’s exit from the EU (Brexit). We used comprehensive trade data from the BACI-CEPII dataset from 1995 to 2022 to create weighted directed graphs to simulate trade ties between countries. Then, using network analysis tools such as sparsification and community recognition, we investigate how these events can modify trade patterns and network architecture. I worked in collaboration with Emanuele Sala, Luca Soleri and Fabio Stefana.

This project for Innovation and Marketing Analytics (Bocconi) focuses on analyzing sentiment and semantic similarity between controversial words regarding the Israel-Palestine conflict. We delved into Reddit comments and articles from three major newspapers, The New York Times, The Guardian, and Al Jazeera, to determine public sentiment and journalists' reporting styles related to the conflict. We used two methods: Word2Vec and sentiment analysis. This dual approach helped us to capture not only how people feel about the topic but also how they conceptually relate different aspects of the conflict through language. I worked in collaboration with Pasquale Caponio, Enrico Cipolla Cipolla and Luca Soleri.

This project for Business Analytics (Bocconi), tests to see whether a blocking policy, blocking people from returning whenever they return an excessive amount, is a policy that reduces the serial returner rate. A survey was run on over 210 correspondents from different walks of life, which showed that the Block Policy would even have an adverse effect to people, returning on average more than when the Block Policy was not introduced. In terms of coding, we performed a sentiment analysis on Amazon reviewsto define which are the reasons behind returning a product. I worked in collaboration with Stefano Bonomi Boseggia, Bumin Kagan Cetin and Ian Ronk.