Building a Vietnamese SentiWordNet Using Vietnamese Electronic Dictionary and String Kernel

摘要

In this paper, we propose a novel approach to construct a Vietnamese SentiWordNet (VSWN), a lexical resource supporting sentiment analysis in Vietnamese. A SentiWordNet is typically generated from WordNet in which each synset has numerical scores to indicate its opinion polarities. However, Vietnamese WordNet is not yet available currently. Therefore, we propose a method to construct a VSWN from a Vietnamese electronic dictionary, not from WordNet. The main drawback of constructing a VSWN from a dictionary is that it is easy to suffer from the sparsity problem, since the glosses in the dictionary are short in general. As a solution to this problem, we adopt a string kernel function which measures the string similarity based on both common contiguous and non-contiguous subsequences. According to our experimental results, first, the use of string kernel outperforms a baseline model which uses the standard bag-of-word kernel. Second, the Vietnamese SentiWordNet is competitive with the English SentiWordNet which uses WordNet when it constructed. All those results prove that our methodology is effective and efficient in constructing a SentiWordNet from an electronic dictionary.