The efficient management and combination of uncertain and conflicting sources of information remain of primal importance for the development of reliable information fusion systems. Advanced fusion systems must deal both with quantitative and qualitative aspects of beliefs expressed by the different sources of information (sensors, expert systems, human reports, etc). This paper introduces the theory of plausible and paradoxical reasoning, known as DSmT (Dezert- Smarandache Theory) in literature, developed originally for dealing with imprecise, uncertain and potentially highly conflicting sources of information providing quantitative beliefs on a given set of possible solutions of a given problem. We also propose in this paper new ideas on a possible extension of DSmT for the combination of uncertain and conflicting qualitative information in order to deal directly with beliefs expressed with linguistic labels instead of numerical values to be closer to the nature of information expressed in natural languages and available directly from human experts.