In this strictly positional paper we propose a general approach to bipolar knowledge representation, where the meaning of concepts can be modelled by examining their decomposition into opposite and neutral categories. In particular, it is the semantic relationship between the opposite categories which suggests the emergence of a paired structure and its associated type of neutrality, being there three general types of neutral categories, namely indeterminacy, ambivalence and conflict. Hence, the key issue consists in identifying the semantic opposition characterizing the meaning of concepts and at the same time the type of neutrality rising in between opposites. Based on this first level of bipolar knowledge representation, paired structures in fact offer the means to characterize a specific bipolar valuation scale depending on the meaning of the concept that has to be verified. In this sense, a paired structure is a standard basic structure that allows learning and building different valuation scales, leading to linear or even more complex valuation scales.