The conceptual representation is one of the most commonly used approaches as a solution for semantic information retrieval. Most approaches apply NLP tools to map terms from queries and documents to concepts and then compute the relevance scores based on the concept representation. However, the mapping results are not perfect due to the erroneous concepts that are generated out of the document context. To overcome this problem, we propose to add a concept selection step in the indexing. Furthermore, we propose in this paper to study the use of semantic similarity distances in the matching step. Then, we propose a method based on adaptive genetic algorithm to combine two SSD.