Evaluating of Word Embeddings Hyper-parameters of the Master Data in Russian-Language Information Systems

Evaluating of word embeddings hyper-parameters for master data quality support task is conducted in this work. We introduce the structure and management of master data. We also describe a method of training the embeddings model for elements and methods for estimating the resulting vectors. Using a corpus of training and validation data sets - 264 thousand records, we have conducted an experiment on models with a different set of parameters and get results. Our vectors give good results in mapping and classification problems on specific industry texts in comparison with standard approaches. In conclusion, we present the main recommendations for the hyper-parameters setting in the task of management of master data for industry conditions. Our methods successfully using in the RPA-systems and in the Data Warehouse like the text analysis module.