PROVING GROUND
about academic work proving ground  

WordNet-based Semantic Similarity

This service is a fast implementation of the Wu & Palmer similarity measure, as described, e.g., in: http://blog.thedigitalgroup.com/:

The Wu & Palmer calculates relatedness by considering the depths of the two synsets in the WordNet taxonomies, along with the depth of the LCS (Least Common Subsumer). The formula is:

score = 2 * depth (lcs) / (depth (s1) + depth (s2)).

WordnNet-distance uses in-memory Jena triplestore with its embedded Lucene library. It requires a specifically compiled out version of the WordNet RDF graph, which is supplied within the GitHub repository.

Try it

word 1:
word 2:


[Compute the similarity]

Semantic similarity and sense disambiguation

As a by-product of the similarity computation, the algorithm effectively disambiguates the meanings of the two words to the ones that maximize the value of the meassure. For instance, the word "shark" has a different meaning when compared to "fish" and a different one when contrasted with "pro" . Conceptually, such process can be described as context-driven, in the sense that the ambiguities are resolved by assuming the closest context (least common subsumer) in which both words have some established meanings.

Additional references