Giuseppe Lancioni

Giuseppe Lancioni

Ph.D XXXIV

Supervisor: Giuseppe Serra

+39 0432 558446

Stanza / Room: Lab AI2

lancioni.giuseppe@spes.uniud.it

Research Project

Text and Graph Generation via Generative Adversarial Networks

Artificial Neural Network (ANN) are computing systems developed with the aim to mimic the behavior of the human brain. They are currently used to for a variety of applications spanning from medical diagnostics to industrial quality control or finance (e.g. automated trading systems) and much more. This activity is well known as Machine Learning or, depending on the complexity of the systems, Deep Learning.

The Generative Adversarial Networks (GAN) are a particular architecture in the field of Artificial Netural Network. They are a promising topic in Artificial Intelligence, as they could be trained to generate fake samples inspired to real ones feeded during training. So far they achieved impressive results in the field of image elaboration and generation.

The aim of my research project is to apply this and others new and powerful tools in the field of Natural Language Processing (NLP), i.e. in the elaboration and generation of texts and graphs.

This could be a great value for the institution I work for, Istituto Nazionale di Statistica (ISTAT, italian stathistical office). It is responsible of the production of the offical stathistic of the country and recently published some experimental stathistics based on the elaboration of great amounts of texts, originated by the web, by mean of Machine Learning techniques.