Two End-to-End Quantum-Inspired Deep Neural Networks for Text Classification

In linguistics, the uncertainty of context due to polysemy is widespread, which attracts much attention. Quantum-inspired complex word embedding based on Hilbert space plays an important role in natural language processing (NLP), fully leverages similarity between quantum states and tokens. A containing multiple meanings could correspond a single particle may exist several possible states, sentence be analogous system where particles interfere with each other. Motivated by quantum-inspired embedding, interpretable complex-valued (ICWE) proposed design two end-to-end deep neural networks (ICWE-QNN CICWE-QNN representing convolutional network ICWE) for binary text classification. They have proven feasibility effectiveness application NLP can solve problem information loss CE-Mix [1] model caused neglecting linguistic features text, since feature extraction presented our learning algorithms, gated recurrent unit (GRU) extracts sequence sentences, attention mechanism makes focus words sentences layer captures local projected matrix. The ICWE-QNN avoid random combination tokens considers textual Experiments conducted five benchmarking classification datasets demonstrate models higher accuracy than compared traditional including CaptionRep BOW, DictRep BOW Paragram-Phrase, they also great performance F1-score. Eespecially, has as well four SST, SUBJ, CR MPQA. It meaningful effictive exploration promote