A New Semantic Similarity Measurement Based on HowNet Concept Tree

This paper puts forward a new depth & path-based semantic similarity method to improve the existing meaning-based approaches in HowNet. Firstly, a complete concept tree is constructed on the sememe tree according to the concept definitions in HowNet. Then an improved depth & path algorithm was put forward, in which five adjustable parameters are used to compute the depth and path of concept in concept tree. This method avoids the computation process of complicated meaning similarity and is more intuitive and efficient. The experiment shows that proposed algorithm has achieved an excellent level comparing with the existing word similarity algorithms.