Feature-based Transportation Sentiment Analysis Using Fuzzy Ontology and SentiWordNet

People are using social media to share their opinions and thoughts about transportation. Sentiment analysis can study these opinions and emotions for the evaluation and improvement of transportation features and services. However, the transportation-related data on social media are unstructured, short length and with a lot of dynamic topics. In addition, the existing systems are discovering sentiments at sentence or document level. These systems are inefficient to extract relevant features, identify polarity orientation of features, and classify the sentiment of features. Therefore, we present a new approach of sentiment analysis for feature extraction and polarity classification. The proposed system is based on fuzzy ontology that presents the relations between concepts semantically in the domain of transportation. The semantic knowledge is employed to identify features in document. The polarity of these features is computed by assigning their opinionated words in document into SentiWordNet. We use logistic regression and multi-layer perceptron along with fuzzy ontology. The experimental results show that fuzzy ontology with learning algorithm is more effective than classifiers without ontology.