An Embedded Tree Matching Algorithm based on Metaphorical Dependency Structure

Knowledge extraction plays an active role in intelligent systems. This paper proposes an embedded tree matching algorithm oriented to knowledge extraction and natural language processing in intelligent systems to extract semantic structures from discourses. The algorithm is designed and tested on Chinese metaphoric expressions to find in a given sentence all the dependency semantic relations that most probably occur in metaphors. The main process is top-down searching and bottom-up amending. Experiment results show the algorithm can expectantly find out accurate dependency relations. This work proposes a preliminary work for discovering potential metaphorical structures or other semantic structures in Chinese sentences which would be useful for intelligent systems capable of natural language understanding.