Big Data re-opened the debate between deterministic and probabilistic methods in linguistics. Semantics seems to be a promising field for their application: Greimas observed how semantic layers are stabilized in a document thanks to redundancy, thus transmitting less information. For example, the context stabilizes one of the possible meanings of ambiguous syntactic structures such as “the chicken is ready to eat.” by reducing the probability of the other meanings. Among different probabilistic algorithms, a model based on quantum geometry could better represent natural language semantics. Semantic relationships seem not commutative [3]; they violates Leibniz's law as it happens in quantum logic; many semantic relationships are not represented by morphosyntactic elements, as the vector state is not observable; the stabilisation of a semantic bind could be modelled by entanglement. We will show how the Bell states can be used to encode the basic fundamental semantic relationships represented by Greimas’s semiotic square, thus providing a clue to develop new algorithms designed for information retrieval.