Traitement quantique des langues : {\'e}tat de l'art

TL;DR

This paper reviews the current state of quantum computing applications in NLP, highlighting approaches that aim to enhance linguistic representations and model performance, and discusses future research directions.

Contribution

It provides a comprehensive overview of quantum NLP methods, including symbolic and hybrid neural approaches, and discusses experimental feasibility and future research perspectives.

Findings

Quantum NLP approaches are feasible for experimental studies.

Hybrid neural and symbolic methods are explored for linguistic phenomena.

Research perspectives include new model design and evaluation methods.

Abstract

This article presents a review of quantum computing research works for Natural Language Processing (NLP). Their goal is to improve the performance of current models, and to provide a better representation of several linguistic phenomena, such as ambiguity and long range dependencies. Several families of approaches are presented, including symbolic diagrammatic approaches, and hybrid neural networks. These works show that experimental studies are already feasible, and open research perspectives on the conception of new models and their evaluation.