We study the global topology of the syntactic and semantic distributional similarity networks for English through the technique of spectral analysis.
Syntax and semantics are two tightly coupled, yet very different properties of any natural language — as if one is from “Mars” and the other from “Venus”.
The syntactic and semantic DSNs are constructed from a raw text corpus.
Spectral analysis refers to the systematic study of the eigenvalues and eigenvectors of a network.
The first eigenvalue tells us to what extent the rows of the adjacency matrix are correlated and therefore, the corresponding eigenvector is not a dimension pointing to any classificatory basis of the words.
Here, we presented some initial investigations into the nature of the syntactic and semantic DSNs through the method of spectral analysis, whereby we could observe that the global topology of the two networks are significantly different in terms of the organization of their natural classes.
See all papers in Proc. ACL 2009 that mention distributional similarity.
See all papers in Proc. ACL that mention distributional similarity.
Back to top.