Artificial Intelligence Algorithms for Natural Language Processing and the Semantic Web Ontology Learning

Evolutionary clustering algorithms have considered as the most popular and\nwidely used evolutionary algorithms for minimising optimisation and practical\nproblems in nearly all fields. In this thesis, a new evolutionary clustering\nalgorithm star (ECA*) is proposed. Additionally, a number of experiments were\nconducted to evaluate ECA* against five state-of-the-art approaches. For this,\n32 heterogeneous and multi-featured datasets were used to examine their\nperformance using internal and external clustering measures, and to measure the\nsensitivity of their performance towards dataset features in the form of\noperational framework. The results indicate that ECA* overcomes its competitive\ntechniques in terms of the ability to find the right clusters. Based on its\nsuperior performance, exploiting and adapting ECA* on the ontology learning had\na vital possibility. In the process of deriving concept hierarchies from\ncorpora, generating formal context may lead to a time-consuming process.\nTherefore, formal context size reduction results in removing uninterested and\nerroneous pairs, taking less time to extract the concept lattice and concept\nhierarchies accordingly. In this premise, this work aims to propose a framework\nto reduce the ambiguity of the formal context of the existing framework using\nan adaptive version of ECA*. In turn, an experiment was conducted by applying\n385 sample corpora from Wikipedia on the two frameworks to examine the\nreduction of formal context size, which leads to yield concept lattice and\nconcept hierarchy. The resulting lattice of formal context was evaluated to the\noriginal one using concept lattice-invariants. Accordingly, the homomorphic\nbetween the two lattices preserves the quality of resulting concept hierarchies\nby 89% in contrast to the basic ones, and the reduced concept lattice inherits\nthe structural relation of the original one.\n