Semantic Similarity Measures

Given the great importance of the calculation of semantic similarity for many computational applications, we observe that a semantic similarity measure should always meet the following requirements: Computationally scalable: This means that it should not be mandatory to process huge amounts of documents to get good results. Reliable: the source, whatever it may be, should be agreed upon by a reliable community. The covering of both general and more specific domains; so that as many requests or queries as possible can be resolved, regardless of the domain in which the calculation of similarity is going to be applied.

As can be seen behind these characteristics, the main choice to be made to solve the problem relies on the data source to be used. In fact, it is necessary to find and process a source of knowledge that covers those requirements in order, on that basis, to draw up a semantic similarity measure that offers results that are close to those that can be achieved with traditional measures.