Semantic Approach to Language Structures Presentation for Machine Learning Algorithms Design

Abstract The problem of establishing transferable language structures is considered. The key idea is developing a synergistic approach combining semantic grammar rules with the machine learning mechanisms of grammar rules extraction from parallel text corpora. The predesigned rules are founded on the unified cognitive structures extracted from the systems of grammar categories of the Russian and English languages and functional roles of language structures in a sentence. Machine learning methods are used to establish the weights of the meaningful language units and structures for probabilistic augmentation of the rule system for syntactic – semantic sentence analysis. The formalism employed for presentation of the English-Russian matches is a unification grammar variant