Lexical sense alignment using weighted bipartite b-matching
Publication Type:
Refereed Conference Meeting Proceeding
Abstract:
In this study, we present a similarity-based approach for lexical sense alignment in WordNet and Wiktionary with a focus on the polysemous items. Our approach relies on semantic textual similarity using features such as string distance metrics and word embeddings, and a graph matching algorithm. Transforming the alignment problem into a bipartite graph matching enables us to apply graph matching algorithms, in particular, weighted bipartite b-matching (WBbM).
Conference Name:
2nd Conference on Language, Data and Knowledge (LDK 2019)
Proceedings:
2nd Conference on Language, Data and Knowledge (LDK 2019)
Digital Object Identifer (DOI):
NA
Publication Date:
20/05/2019
Conference Location:
Germany
Research Group:
Institution:
National University of Ireland, Galway (NUIG)
Open access repository:
Yes