Student

AMIR ALI MALEKANI NEZHAD

Supervisor

Aznul Qalid Md Sabri

Collaborator

Qualition

This paper is dedicated to a novel approach to Quantum Natural Language Processing as a translation of Neural Network models onto the Quantum Computing paradigm by representing the core model using quantum parameterized circuits whilst not assuming any quantum mechanical nature for language and/or NLP, motivated by the potential speedups offered by quantum computing.

Through this paper we will go through the motivation for this approach, the theoretical approach taken, the detail of the pipeline, experiments with different parsers and ansatzes and lastly the comparison between the two paradigms in a manner to compare accuracy and computational speed. Furthermore, we will present results on the QNLP model conducted on a Noisy IntermediateScale Quantum (NISQ) simulator (TKET Model) and Ideal Simulator (NumPy Model) for datasets of size ≥ 300 sentences. By utilizing the formal similarity(representation similarity) of the “compositional model of meaning” by Coecke et al. (2010) with quantum theory, we will provide representations for sentences which possess a natural mapping to quantum circuits. We will utilize these representations to implement and successfully train a QNLP model that will solve a binary sentence classification task for depression detection with high accuracy and low loss.

Keywords : Quantum Natural Language Processing, Quantum Machine Learning, NISQ, Binary Classification, Depression Detection, Compositional Model of Meaning