Sentence Comprehension and Semantic Syntheses by Cognitive Machine Learning

Abstract

Recent development in machine learning and computational linguistics has enabled cognitive machines to understand the semantics of human expressions. A system for sentence syntactic analysis and semantic synthesis is developed based on denotational mathematics. Machine sentence learning and comprehension are reduced to the building of a composed concept that maps the semantics of the subject onto the counterpart of object(s) represented by formal concepts and phrases. A set of semantic operations such as concept composition, modification, generalization, specification, extension and reduction is formally specified based on concept algebra and semantic algebra for machine learning. An Algorithm for Unsupervised Sentence Learning (AUSL) is designed and implemented, which expresses a learnt sentence as a knowledge graph related to the semantic hierarchy of the machine's knowledge base. Experimental results demonstrate the autonomous learning algorithm and case studies on machine learning towards applications in cognitive robots and knowledge learning systems.