Contributo in atti di convegno, 2019, ENG, 10.1007/978-3-030-36599-8_16

Applying predictive models to support skos:ExactMatch validation

Riccardo Albertoni

Istituto di Matematica Applicata e Tecnologie Informatiche "Enrico Magenes", Consiglio Nazionale delle Ricerche (IMATI-CNR), Via De Marini, 6, 16149 Genova, Italy

The paper investigates the use of Machine Learning (ML) to support experts validating skos:exactMatch links. It trains ML techniques provided by RapidMiner with manually validated links and shows how to use the obtained predictive models for saving expert efforts. The obtained results are preliminary but encouraging: the trained predictive models reduce up to 70% the number of manual checking required from experts, leaving only 10% of the wrong links unnoticed. Cutting the 70% of the expert burden is crucial, especially when dealing with the validation of large sets of links.

13th International Conference on Metadata and Semantic Research (MTSR 2019), pp. 187–193, Rome, Italy, 28-31/10/2019

Keywords

Linkset correctness, Quality, Expert validation, Predictive Models

CNR authors

Albertoni Riccardo

CNR institutes

IMATI – Istituto di matematica applicata e tecnologie informatiche "Enrico Magenes"