Возможности нейросети для поиска когнатов для установления новых этимологий и источников заимствований в шира-югурском языке

Cognate Identification Neural Network and Its Capabilities to Establish New Etymologies and Borrowing Sources in Eastern Yugur

Introduction. The paper describes some results obtained from a cognate identification neural network designed to establish new etymologies and sources of borrowings in Eastern Yugur. Materials and methods. The study provides an overview of existing neural network models, results of their work, and characterizes the available dictionaries of Eastern Yugur. The latter’s etymologies have been specified on the basis of Mongolic-language dictionaries uploaded onto the LingvoDoc platform. The work employs the comparative historical method and certain functional tools of the platform that have proved instrumental in identifying cognates for a number of Eastern Yugur words and reconstructing some essentials of Proto-Mongolic. Results. The article describes the principles of the neural network that follows the Siamese pattern and consists of two identical branches. A total of 40 Proto-Mongolic reconstructions — previously known only for the North Mongolic languages — have been implemented. In addition, the paper introduces 11 examples of early Chinese borrowings to Eastern Yugur, since those are available in other Mongolic languages. A number of Proto-Mongolic lexical reconstructions dealing with material culture are noteworthy enough: *(h)iliɣür ‘[press] iron’, *kükür ‘sulfur’, *jaŋ-‘cement’, *kas ‘jasper, jade’, *kuruɣub-‘thimble’. Efforts aimed at supplementing existing etymologies with data on Eastern Yugur and — in some cases — those from dictionaries of other Mongolic languages available on the LingvoDoc (Classical Mongolian, Mongolian, Buryat, Oirat, Dagur, Dongxiang, Bonan) and verifying reconstructed lexemes through Chinese dictionaries for borrowings make it possible to deepen our knowledge of Mongol cultural history and even specify sources of certain inventions. © KalmSC RAS, 2025 and Normanskaya J. V., Goncharova O. V., Kukanova V. V., Chushkaeva Z. I., 2025.