AUTOMATION OF BEER PROFILE DESCRIPTION USING NATURAL LANGUAGE PROCESSING AND SENSORY DATA

This article presents a scientifically grounded approach to automating the description of beer sensory profiles using modern Natural Language Processing (NLP) methods and the analysis of sensory data obtained from professional tastings. In the context of the growing digitalization of production and marketing processes in the brewing industry, there is an increasing demand for standardized, objective, and reproducible methods for capturing and presenting organoleptic product characteristics. The development of such methods is relevant both for large-scale breweries and small craft beer producers striving for consistent quality and accurate communication of product features to consumers. The aim of this study is to develop an intelligent system based on neural network architectures, in particular the Bidirectional Encoder Representations from Transformers (BERT) model, capable of automatically processing unstructured tasting texts and transforming them into structured descriptions that meet professional standards. The use of pre-trained language models enables the efficient extraction of descriptors, classification of flavor and aroma characteristics, and the generation of textual product profiles suitable for both internal quality control and external marketing purposes. The research details the stages of dataset collection, preprocessing, annotation of key features, formalization of sensory categories, selection of optimal model parameters, and evaluation metrics for text generation quality (such as BLEU, ROUGE, etc.). The tested prototypes demonstrated a high degree of alignment between generated descriptions and expert evaluations, confirmed through both manual and automated assessment. The proposed system can be integrated into corporate information platforms, including ERP and CRM systems used by brewing companies, and applied in scientific or research projects related to the qualitative evaluation of food products. The results confirm the effectiveness of NLP and artificial intelligence technologies in sensory analysis tasks and highlight the prospects for the further application of such approaches in related sectors of the food industry.