In previous work, we presented an approach for link prediction using a probabilistic description logic, named crALC. Inference in crALC, considering all the social network individuals, was used for suggesting or not a link. Despite the preliminary experiments have shown the potential of the approach, it seems unsuitable for real world scenarios, since in the presence of a social network with many individuals and evidences about them, the inference was unfeasible. Therefore, we extended our approach through the consideration of graph-based features to reduce the space of individuals used in inference. In this paper, we evaluate empirically this modification comparing it with standard proposals. It was possible to verify that this strategy does not decrease the quality of the results and makes the approach scalable.
|Idioma original||Inglés estadounidense|
|Número de páginas||12|
|Estado||Publicada - 1 dic 2012|
|Publicado de forma externa||Sí|
|Evento||CEUR Workshop Proceedings - |
Duración: 1 ene 2016 → …
|Conferencia||CEUR Workshop Proceedings|
|Período||1/01/16 → …|
Ochoa Luna, J. E., Revoredo, K., & Cozman, F. G. (2012). An experimental evaluation of a scalable probabilistic description logic approach for semantic link prediction. 63-74. Papel presentado en CEUR Workshop Proceedings, .