My main research interests deal with Information Retrieval and how to adapt Information Retrieval Evaluation to user signals via stochastic models.
The main research contributions of my PhD thesis are:
Designing of a formal framework for IR evaluation measures and analysis of the properties characterizing the different IR measures;
A family of new evaluation measures (called Markov Precision) built on top of Markov-chain based user models. This evaluation metric can be calibrated on the user dynamic and takes into account the user experience;
Adapting click models for the specific case of job search (Use case: SEEK Job search engine) by dealing with data sparsity;
Integrate the user dynamic defined with a Markovian model in LambdaMART to improve learning to rank;
A novel methodology for dealing with multiple crowd assessors, who may be contradictory and/or noisy (Referred as AWARE-Assessor-driven Weighted Averages for Retrieval Evaluation).
The MATlab Toolkit for Evaluation of information Retrieval Systems (MATTERS) has been updated with new measures and analysis functions: matters.dei.unipd.it
I organized the 1st International Workshop on LEARning Next gEneration Rankers, co-located with ICTIR 2017, held in Amsterdam, October 1, 2017: learner2017.dei.unipd.it
I contributed to the organization of the 38th European Conference on Information Retrieval held in Padua, from March 20 to March 23, 2016: ecir2016.dei.unipd.it