Natural language understanding and communication is a key aspect of human-robot collaboration. Concerning natural language understanding, various features need to be covered for creating an overall system. Apart from semantic parsing and mapping, the reusability and extendability of such a system requires consideration. In this thesis a system for natural language understanding based on dependency parsing is designed. The focus lies on extracting the meaning of spoken commands. A communication component is added for enabling on the one hand ambiguity resolving and on the other hand task suggestions on the part of the robot. The proposed system is evaluated with respect to its natural language understanding capabilities. The evaluation revealed the high impact of speech recognition on semantic parsing. The significantly enhanced spoken command understanding unveiled the benefit of an ambiguity resolving communication component.