The lexical knowledge and semantic representation of E-HowNet

E-HowNet,extended from HowNet,is a frame-based entity-relation model aimed for the goal of natural language understanding. We adopt the taxonomy structure of sememes of HowNet and modify its lexical semantic representation to suit for automatic semantic composition and decomposition. In E-HowNet, all lexical senses, including function words and content words,are expressed by either semantic primitives ( called sememes in HowNet) or basic concepts and their semantic relations in a uniform way. Therefore senses of phrases or sentences can be derived by unifying the lexical sense representations of their words. The consistency of E-HowNet is examined by checking the taxonomy of synonym classes which is automatically constructed according to E-HowNet lexical expressions. Difficulties in lexical semantic representation are also encountered and solutions to these problems are offered. E-HowNet systematizes the complex nature of language semantics and hopefully may shed light on the goal of natural language understanding.