The construction of a sentiment dictionary for public opinion mining on social media: a study on nuclear power opinion on Twitter

摘要

The opinion mining process includes the classification of a set of documents, whether they are positive, negative or neutral. To support the process, it is important to construct and utilize a sentiment dictionary which defines sentimental polarity of major vocabularies. Topics that have been discussed on social media such as Twitter and online communities are changing frequently and rapidly in real time. Therefore, the effective construction and evolution in planning of a sentiment dictionary is essential to improving performance of opinion mining on social media. This paper aims to propose a method to construct and evolve a sentiment dictionary reflecting changes of public issues which are discussed in online. Through an experiment using tweets on nuclear power on Twitter, we can confirm that the proposed approach can contribute to improvement in classification accuracy.