Natural language sentences can be represented as vectors in a high dimensional vector space. Generally, these models are based on bag of words approaches, and therefore they do not fully capture the semantics of sentences which depends both by the semantics of the words, and their order in in the phrase. In this work we propose a sub-symbolic methodology to encode natural language sentences considering both these two aspects. The proposed approach exploits the properties of Geometric Algebra rotation operators, called rotors, to code sentences through the rotation of an orthogonal basis of a semantic space. The methodology is based on three main steps: the construction of a semantic space, the association of ad-hoc rotors to sentence bigrams, and finally the coding of the sentence through the application of the obtained rotors to a standard basis in the semantic space.