International Journal of Computational Linguistics Research Mohcine Madkour, Driss Elghanami, Abdelilah Maach, Hasbi abderrahim 3 3 2012 http://www.dline.info/jcl/fulltext/v3n3/4.pdf In ubiquitous computing the discovery process may retrieve many services when in fact only one of them fit exactly user satisfaction, besides, after discovering is achieved, there is a set of candidate services between theme a selection must be made. In fact, discovery is a prerequisite for selection, but selection is the main purpose. Actually, uncertainty of context information may lead to inexact matching between already discovered and required service capabilities, and conse-quently to the non selection of fitting services. In order to handle incomplete context information, we propose in this paper a workflow-based algorithm allowing inexact matches for matching contextual service descriptions using similarity measures. Service description and request are compared using four kinds of similarity measures: syntactic, linguistic, structural and QoS semantic measures, which compare individually requested and provided properties represented as workflow nodes, and thereafter the global measures which take into account context and service as a whole are aggregated by means of the linguistic quantifier “almost all”. In our approach we consider that functional aspects of a service are already met and we focus on non-functional and QoS-related aspects of service description to rank-order the discovered services