Quantum Entanglement Based Sentence Similarity Computation

Sentence representation is one of the foundations of natural language processing. A quantum entanglement-based approach is provided to determine the sentence similarity. The sentence embedding based on quantum computation considers the semantic structure of a sentence, tensor product and attentive mechanism. Main components of modifiers of a complete sentence are obtained by using the attentive mechanism combining with quantum entanglement of two consecutive notional words. The semantic features and syntactic structures of sentences are extracted by the tensor product of two consecutive notional words without any big-corpus or prior knowledge. Experimental results implemented on 17 datasets with an excellent effect show that taking the quantum entanglement into the sentence embedding yields improvements on mining semantic relations and syntactic structures. These results shed new light on sentence representation with limited resources, and represent a new step towards the quantum entanglement-based application in natural language processing.