TEXT CLASSIFIER USING A HYBRID APPROACH: CLASSICAL DATA PREPROCESSING AND QUANTUM SVM MODELS

Babasanova N. S., Kanev A. I., Mikheeva V. A.

The work addresses the problem of the lack of practical implementations of quantum algorithms for natural language processing (NLP). The goal of the work is to present an approach to building a quantum text classifier for the binary classification of SMS messages with a split in computations: data preprocessing is performed on a classical computer, while the classification stage is carried out using a quantum algorithm. A comparison with a classical SVM (scikit-learn) was conducted for the quantum PegasosQSVC and QSVC algorithms; with identical data preprocessing, the quantum algorithms demonstrate comparable accuracy: 94% for the classical SVC and 96% for PegasosQSVC (on the full dataset), and 97% for SVC and 96% for QSVC (on a reduced dataset).

Keywords:machine learning, NLP, quantum computing, SVM, quantum machine learning, Qiskit.