Determining Semantic Similarity of Behaviour Using Natural Semantic Metalanguage to Match User Objectives to Available Web Services

Abstract

Natural Semantic Metalanguage specifies a set of semantic primitives identified by linguists since the 1970s as being present in all analysed languages and not capable of further reduction. In this article, the 63 semantic primitives are used to define the semantics of user objectives and web services in the form of semantic explications, which are then compared to determine whether the web services are likely to be helpful in meeting the user objective. The comparison of the user objectives and web services is a two stage process. Firstly, the content is compared by classifying the semantic primitives from the candidate web service and user objective on the basis of whether the primitives are common or similar. On the basis of these classifications, the percentage match and semantic relationship (subset, superset, overlaps, disjoint, identical) are determined. Secondly, the order of the semantic primitives is compared and the edit distance determined as a measure of semantic similarity. The method is tested using two examples: a comparison of spatial relations and a comparison of a user objective and three geospatial web services. The results show that the method is able to determine which concepts are broadly semantically similar and which are not.