Many attempts have been made on processing natural queries on relational databases, because this feature should render users virtually no boundary on expression search requests. Recently, research trends on this field have moved from compositional semantics into SQL-oriented processing on natural queries. However, this approach still suffers the problem of imprecise and incomplete information which often occurs when users submit their queries casually in practical situations. In this paper, we show an approach on translating the natural queries into conceptual graphs using semantic information captured by a domain ontology. Due to the well-structured representation of conceptual graph, we can resolve the impreciseness and retrieve the incompleteness when occurring. Experimental results have shown that our approach is quite promising.