In this paper, we study the problem of structural sense ranking for tree data using a multi-relational PageRank approach. By considering multiple types of structural relations, the original tree structural context is better leveraged and hence used to improve the ranking of the senses associated to the tree elements. Upon this intuition, we advance research on the application of PageRank-style methods to semantic graphs inferred from semistructured/plain text data by developing the first PageRank-based formulations that exploit heterogeneity of links to address the problem of structural sense ranking in tree data. Experiments on a large real-world benchmark have confirmed the performance improvement hypothesis of our proposed multi-relational approach.
Multi-relational PageRank for Tree Structure Sense Ranking
TAGARELLI, Andrea
2015-01-01
Abstract
In this paper, we study the problem of structural sense ranking for tree data using a multi-relational PageRank approach. By considering multiple types of structural relations, the original tree structural context is better leveraged and hence used to improve the ranking of the senses associated to the tree elements. Upon this intuition, we advance research on the application of PageRank-style methods to semantic graphs inferred from semistructured/plain text data by developing the first PageRank-based formulations that exploit heterogeneity of links to address the problem of structural sense ranking in tree data. Experiments on a large real-world benchmark have confirmed the performance improvement hypothesis of our proposed multi-relational approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.