FIELD: computer technology.
SUBSTANCE: invention relates to the field of computer technology for identifying blocks of related words in documents using neural networks. A set of document words is obtained, the document has the first block of related words, a set of vectors representing the set of words is determined, the set of vectors is processed, using the first neural network to obtain a set of recalculated vectors having values based on the set of vectors, a set of connectedness values corresponding to connections between two words in the document is determined, and, using the set of recalculated vectors and the set of connectedness values, the first block of related symbol sequences is identified.
EFFECT: more efficient detection of blocks of related words in documents of complex structure, thereby increasing both the accuracy of identification and the processing speed of the computing device.
20 cl, 13 dwg
