Hybrid Quantum-Classical RNN for NLP

• Developed a novel Quantum Recurrent Neural Network (QRNN) architecture that combines classical

embeddings and control networks with parameterized quantum circuits for sentiment analysis

• Implemented parameter-shift rule gradient computation to make quantum measurements differentiable,

enabling backpropagation through quantum circuit parameters

• Built complete training pipeline: TF-IDF vocabulary filtering, GloVe embeddings, classical feedforward controller

generating rotation angles, 4-qubit variational circuit with entangling gates, and Pauli measurement readouts

• Architected quantum state evolution system where previous quantum state serves as input to next timestep,

creating quantum memory analogous to classical RNN hidden states

• Tools Used: PyTorch, Qiskit, Quantum Computing, Deep Learning