Title: A lukewarm recommender system for Twitter
Speaker: Fernando Alves, LASIGE – DI/FCUL
When: March 22 (Thursday) at 12:00
Recommender systems suffer from the so-called “cold-start problem”, where the system requires a considerable amount of data to properly recommend items to users. In this seminar, I will present a “lukewarm” recommender (since it requires few samples) to tackle this issue. The user presents a set of topics of interest, each represented by ten tweets. These tweets are used to train a Bi-Term Topic Model and an autoencoder ensemble. We show this approach works in a set of cybersecurity topics.