Conceptual Search Based on Semantic Relatedness

Traditional search engines based on syntactic search are unable to solve key issues like synonymy and polysemy. Solving these issues leads to the invention of the semantic web. The semantic search engines indeed overcome these issues. Nowadays the most important part of the data remains unstructured documents. It is consequently very time consuming to annotate such big data. Concept based retrieval systems intend to manage directly unstructured documents. Semantic relationships are their main feature to extend syntactic search. In most of the methods implemented so far, concepts are used for both indexing and searching. Words remain the smallest unit to process semantic relatedness. The differences persist in the way that concepts are represented, mapped to each other, and managed for the sake of indexing and/or searching. Our approach is based on Wikipedia concepts. Concepts are represented as an undirected graph. Their semantic relatedness is computed with a distance derived from a semantic similarity measure. The same distance is used to calculate both semantic relatedness and query matching.