Perceiving Service Quality from Social Media via Natural Language Processing Model

The prosperity of social media provides us with a vast repository of knowledge. Mining users' perceptions of a specific service from social media can help service providers improve their service quality and identify potential issues that undermine the quality. However, mining user opinions from social media is still challenging. First, social media data are highly unstructured. Second, a sentence may mention multiple objects with different attitudes. Third, to analyze the factors that impact service qualities, it is critical to identify both mentioned objects and corresponding attitudes towards each of them. To tackle the above obstacles, this paper proposed a novel framework based on natural language processing (NLP). Specifically, our framework can extract fine-grained opinions hidden in each piece of social media data. In this paper, an opinion is a triple consisting of three elements: target, aspect, and sentiment. A sentiment expresses a user's attitude about the service, e.g., positive, negative, or neutral. A target is the element that causes the sentiment, i.e., the reason. An aspect is the class of a target. Based on such triples, users' perceptions of service quality can be measured precisely. To empirically study the effectiveness of our framework, we conduct an experiment with Macau International Airport (MIA) and Twitter. The results demonstrate that our framework can successfully perceive user opinions from social media to help improve the service quality of MIA.