Nuova ricerca


Dottorando presso: Dipartimento di Ingegneria "Enzo Ferrari"

Home | Curriculum(pdf) |


2021 - Progressive Query-Driven Entity Resolution [Relazione in Atti di Convegno]
Zecchini, Luca

Entity Resolution (ER) aims to detect in a dirty dataset the records that refer to the same real-world entity, playing a fundamental role in data cleaning and integration tasks. Often, a data scientist is only interested in a portion of the dataset (e.g., data exploration); this interest can be expressed through a query. The traditional batch approach is far from optimal, since it requires to perform ER on the whole dataset before executing a query on its cleaned version, performing a huge number of useless comparisons. This causes a waste of time, resources and money. Proposed solutions to this problem follow a query-driven approach (perform ER only on the useful data) or a progressive one (the entities in the result are emitted as soon as they are solved), but these two aspects have never been reconciled. This paper introduces BrewER framework, which allows to execute clean queries on dirty datasets in a query-driven and progressive way, thanks to a preliminary filtering and an iteratively managed sorted list that defines emission priority. Early results obtained by first BrewER prototype on real-world datasets from different domains confirm the benefits of this combined solution, paving the way for a new and more comprehensive approach to ER.

2021 - The Case for Multi-task Active Learning Entity Resolution [Relazione in Atti di Convegno]
Simonini, Giovanni; Saccani, Henrique; Gagliardelli, Luca; Zecchini, Luca; Beneventano, Domenico; Bergamaschi, Sonia

2020 - Entity resolution on camera records without machine learning [Relazione in Atti di Convegno]
Zecchini, L.; Simonini, G.; Bergamaschi, S.

This paper reports the runner-up solution to the ACM SIGMOD 2020 programming contest, whose target was to identify the specifications (i.e., records) collected across 24 e-commerce data sources that refer to the same real-world entities. First, we investigate the machine learning (ML) approach, but surprisingly find that existing state-of-the-art ML-based methods fall short in such a context-not reaching 0.49 F-score. Then, we propose an efficient solution that exploits annotated lists and regular expressions generated by humans that reaches a 0.99 F-score. In our experience, our approach was not more expensive than the dataset labeling of match/non-match pairs required by ML-based methods, in terms of human efforts.