My main research interests concern Information Retrieval System Evaluation and predictive methods applied to IRS Evaluation.

My research work has been focused on performance prediction for Information Retrieval Systems:
In many engineering fields, it is possible to mathematically and statistically compute ahead the expected behaviour and performance of a system so that it can be designed and developed to achieve such performance. For example, a bridge is built keeping in mind the expected weight it will have to endure. This important feature is missing in systems where the human cognitive effort is at stake, such as Information Retrieval (IR): the human component makes hard to predict the performance achieved by IR systems before they are actually developed, deployed, and tested. Thus evaluation procedures require expensive user studies or are applied on toy problems and might lack generalization capability when the system is used in real-world scenarios. Yet, the research reached a consensus on the fact that it is possible to develop models to gain insight on the system's performance. With my research work, we aim at identifying and develop models capable of bounding and predicting the IR system performance, without relying on a post-hoc evaluation phase. Secondly, we plan to identify features of collections and data sets, tasks, users, and models, that can be efficiently gathered and effectively leveraged by the predictive model.

• Information Retrieval Systems
• IRS Evaluation
• Performance Prediction
• Machine Learing
• Recommender Systems