Latent Semantic Kernels for WordNet: Transforming a Tree-Like Structure into a Matrix

WordNet is one of the most widely used linguistic resources in the computational linguistics society. However, many applications using the WordNet hierarchical structure are suffering from the word sense disambiguation (WSD) caused by its polysemy. In order to solve the problem, we propose a matrix representing the WordNet hierarchical structure. Firstly, we transform a term as a vector with elements of each corresponding to a synset of WordNet. Then, with singular value decomposition (SVD), we reduce the dimension size of the vector to represent the latent semantic structure. For evaluation, we implement an automatic assessment system for short essays and acquire reliable accuracy. As a result, the scores which are assessed by the automatic assessment system are significantly correlated with those of human assessors. The new WordNet is expected to be easily combined with other matrix-based approaches.