Automation of Numerals Form Replacement Process

Text-to-speech systems usually require numbers in text to be written as words rather than digits. the research presented in the paper is aimed at developing an automatic programming solution for replacing a representation of numbers from digit to word form. the study used ordinal and cardinal numbers from texts in the Russian language. Two different approaches for replacement were applied. the first one implies that it is possible to create certain language processing rules that could cover all of the linguistic patterns. Therefore, this solution must take into account prepositions, and the fact that the word form of a numeral depends on many other syntactic constructions. the second one relies on machine learning methods, where sentences with numbers represented by digits are used as inputs. Aside from model training and fine-tuning, this method includes collecting data in the form of sentences in Russian, which contain different kinds of numbers, such as integer and real numbers, ranges, dates and times. the paper shows the results of testing the developed models.