Abstract: Ontology learning is a hot research field in semantic technology and its application, and the domain-specific concept sieving is the foundation of ontology learning. Although the method based on domain relevance expressions and domain consensus expressions displays the good effectiveness for the domain-specific concept sieving, it exists the faults of unilateral description information. So this paper presented an improved sieving algorithm of the domain-specific concept to solve the above problems. First, low freduency with synonymy and the part of relationship words set were identified and redundant concepts were filtered out through calculating the semantic similarity between candidate concept, and then using the improved field concept similarity and domain concepts consistent degree formula sieve concepts. The experiments show that this method improves the effectiveness of the domairrspecific concept sieving.

Key words: Semantic technology, Ontology learning, Domain-specific concept, Sieving algorithm, Context