Abstract-Concept Description Language (CDL) is a common language that represents the semantics of content in a simple and structured manner. In particular, it is intended to describe Natural Language (NL) texts in a format that can be understood and processed by computers. Since words with multiple meanings can be found from texts, it becomes necessary to perform Word Sense Disambiguation (WSD) in order to achieve a correct representation. This paper presents a WSD approach that determines best candidates for word meanings and contributes to a semi-automatic conversion of NL into CDL. We perform preliminary experiments by evaluating the approach with some test sentences and comparing with other WSD methods. Results suggest that the existence of a proper correspondence of syntactic and semantic relations for the WSD process may lead to an accurate conversion to CDL.