Syntactic Model for Russian: Deep Learning Models with Dependency Parsing Scheme

This work presents results of effectiveness analyses of using deep neural networks models for syntactic parsing of SynTagRus dataset. A set of modular neural network topologies based on composition of Stack long short-term memory layers, multilayer perceptron has been compared with widely used algorithms of classification on base of Gradient boosting trees and Support vector machine. Results allow to select set of state-of-art neural networks model and estimate their perspective.