Chinese word semantic relation classification based on multiple knowledge resources

Abstract

Chinese word semantic relation classification is an important and challenging task in the field of natural language processing. This paper describes our method to classify Chinese word semantic relation based on multiple knowledge resources at NLPCC Evaluation. Firstly, given pairs of Chinese words, we try to utilize different knowledge resources, such as Tongyici Cilin and HowNet, to classify them into four kinds of semantic relations, which are synonym, antonym, hyponym and meronym. Secondly, for those uncovered pairs of Chinese words, we translate them into English, then classify them with the help of English knowledge resources, such as WordNet and BabelNet. Experiments on the evaluation dataset at NLPCC 2017 demonstrate that the method can achieve the macro-averaged F1-Score of 0.634 and precision of 0.875. Among all of the participants, the method get the best precision, which shows its superiority over other methods on precision.