Computing Chinese Semantic Orientation Via Distributional Similarity
Huixian Li · Xi'an Jiaotong Daxue xuebao · 2009
An algorithm for Chinese semantic orientation calculation that uses distribution similarity is proposed to solve the problem that existing methods take less implied semantic into consideration in semantic orientation inference.The Chinese semantic orientation calculation is carried out in two steps.The first step calculates the distribution similarities using dependency grammar analysis and statistical tools.HowNet and Chinese conjunction features are introduced in semantic similarity calculation to optimize the corpus-based statistical results.The second step adopts an undirected weighted graph clustering algorithm to infer semantic orientation.Because it is an NP-hard problem to obtain the optimal clustering solution,a greedy algorithm is used to get an approximate solution.Experiments on the testing corpus show that the accuracy of the proposed algorithm is 80% and is better than both the corpus-based statistic algorithm and the HowNet-based algorithm.The results demonstrate that the proposed method is feasible and effective to improve the accuracy of Chinese semantic orientation calculation.