I know WordNet is capable of providing semantic distances between two words, but how about phrases containing more than one word?
Is there any open libraries which allow me to measure the semantic distance between two different phrases?
I know WordNet is capable of providing semantic distances between two words, but how about phrases containing more than one word?
Is there any open libraries which allow me to measure the semantic distance between two different phrases?
Linked to semantic similarity (distance) between words is the notion of similarity between concepts. A common approach is to extract concepts expressed in texts/phrases in order to reduce bags of words to bag of concepts. Such a process is not trivial (e.g. polysemy,ambiguity) but domain specific tools, such as MetaMap for the biomedical domain, can be used to map unstructured texts to unambiguous concepts/terms defined in an ontology/terminology.
Such bags of concepts can therefore be used to assess the similarity between texts/phrases taking advantage of semantic graphs (e.g. ontologies, WordNet) which is interesting to capture their real meaning. The Semantic Measures Library (SML) can be used to estimate the similarity or distance between such bags of concepts considering a semantic graph structuring the concepts. The library is developped in Java, open source and implements a wide range of measures. A toolkit is also provided to compute semantic similarities using a command line interface.
You can download the SML and the toolkit from the dedicated website http//www.semantic-measures-library.org
Hope these help