Web pages of R. Bernardi at UniTN

Research Interests

Through the years, since I started my research career in 1997, I have got involved in both theoretical and applied research projects focused mainly on: Language and Vision, Natural Language Syntax-Semantics Interface and Natural Language Interface to Information Systems

Language and Vision

I started working on language and vision models in 2011 driven by the interest in studying the interaction between language and vision inputs in computational semantic tasks traditionally thought only for natural language representations. I carried out my first work on LV by supervising Thu Dieu Le in collaboration with Jasper Uijlings. In the more recent past, I have been working on foundational issues -- the acquistion of vague expressions, like quantifiers (see Pezzelle's PhD Thesis), and on interactive multimodal agents (see Shekhar's, Greco's and Testoni's PhD theses). I am the coordinator of the Language-Vision group.

Natural Language Syntax-Semantic Interface

My main interest is on the investigation of formal principles behind natural language. To this end, I carry out research in different directions both at the theoretical and applied level, usually working with logic based frameworks tracing back to the Montague Grammar tradition.

I have worked with Marco Baroni and Roberto Zamparelli on Compositional Distributional Semantics Model as part of the ERC COMPOSES team.

On the logic side, I've collaborated with Carlos Areces, Rajeev Goré, Michael Moortgat. We've investigated logical principles of the Lambek Calculus via Display Logic in order to extend the expressivity of the Logic. With Michael Moortgat, we have worked on the Dual Lambek Calculus exploring the connection with Continuation Passing Style Semantics. On the linguistic side, I've worked with Anna Szabolcsi and Chris Barker on formal semantics aspects challenging for formal grammar systems.

Natural Language Interface to Information Systems

While part of the KRDB, I have learned about DB, KR and ontologies and got interested in how to access them via natural language. Again, I've tackled this problem from both a theoretical and data-oriented perspective.

Together with Diego Calvanese and Camilo Thorne, we've worked on Controlled Natural Langauges. We have investigated fragments of natural language suitable for querying and updating ontologies.

I am interested in Interactive Question Answering. I have started working in this direction on QA within Maarten de Rijke's group. I have worked with Manuel Kirschner and Massimo Poesio. We have collected corpora of dialogues and studied their properties so to build an interactive question answering system that helps the user extract the needed information. We took the FAQs of the FUB Library as domain of investigation. See BoB for further details.

Together with Diego Calvanese and Barbara Plank I have worked on a the MuSiL project (Multilingual Access to Library Catalogues) carried out in collaboration with CELI and the Library of FUB. MuSiL has been the starting point for the EU project CACAO of which I have been FUB PI. A spin-off, CLS, has been launched in November 2010 based on the CACAO software.