Concept-based lexical-semantic unsupervised learning system
V I Filatov · Journal of Physics Conference Series · 2014
Concept learning is essential for automatic knowledge acquisition and consists in linking semantic and linguistic data together. In this work, a system of concept learning, which mimics to some extent human infant learning, is presented. This system performs visual and audial feature extraction and construction of concept database by maximizing of their mutual information. Experiments show 90% recognition rate of learnt concepts.