FIELD: information technology.
SUBSTANCE: invention relates to the field of information technologies, specifically to text recognition systems. Method of analysing unstructured documents presented in a digital format, comprising: obtaining a document dummy comprising a plurality of fields; identification of document, determination of multiple sequences of symbols in document, processing by means of processor device of multiple sequences of symbols in document using first neural network associated with first type of documents, to determine relationship between first field of plurality of fields with first sequence of symbols from plurality of sequences of symbols in document, and wherein said first neural network comprises one or more neural subnets containing at least one hidden layer of neurons.
EFFECT: technical result is aimed at improvement of accuracy of field recognition in unstructured documents.
19 cl, 16 dwg
