A Few Shot Learning Scheme for Quantum Natural Language Processing

TL;DR

This paper introduces a framework for Few Shot Learning in Quantum Natural Language Processing, combining classical training and variational quantum training to improve efficiency and reduce errors in quantum NLP tasks.

Contribution

It proposes a novel hybrid quantum-classical approach that enhances quantum NLP by dividing encoding into classical and variational parts, addressing high costs and error susceptibility.

Findings

Framework effectively leverages classical data to reduce quantum processing load

Demonstrates improved performance in quantum NLP tasks with limited quantum resources

Explores the potential of hybrid quantum-classical models in natural language understanding

Abstract

The field of Quantum Computation is plagued by issues that limit the implementation and development of quantum systems and quantum algorithms. Issues which force the development of Hybrid Quantum-Classical algorithms, such as the quantum DisCoCat implementation for Natural Language Processing. These require a high processing cost and are susceptible to errors due to Out of Vocabulary words. In this work, we develop a framework to implement Few Shot Learning for Quantum Natural Language Processing, by modifying the encoding ans\"atze and dividing it into two parts, the first one leveraging the vast corpus of classical training already available, and the second variationally training on the task. This framework is then put to the test to explore its behaviour and its power in extracting as much useful work from each call to a quantum system as possible.