<p>This article is devoted to solving the problem of forming quantum features and classifying texts based on them. A quantum one-hot algorithm is proposed for classifying sarcastic expressions using a neural network. Today, sarcasm is frequently encountered on social networks, information websites, and messengers. Its analysis and classification constitute one of the pressing challenges in the field of natural language processing. Sarcastic expressions are generally context-dependent. To detect them, the proposed quantum one-hot algorithm was applied to represent textual data in vector form, and a neural network was trained on the generated vectors. Based on the quantum one-hot algorithm, textual data <br /> are transformed into quantum amplitudes, after which measurement is performed according to the vectorization results. Through measurement, the quantum data are converted into classical form. These data are then used to train an LSTM neural network. Experimental results demonstrated that quantum approaches outperform classical ones. This study highlights the potential for applying quantum encoding methods in deep learning systems, which may serve as a foundation for new approaches in text analysis.</p>