Sentiment Analysis of Transportation Using Word Embedding and LDA Approaches

초록

Intelligent Transportation System (ITS) is facing technical challenges: extracting valuable information from social networking and the representation of extracted information for the purpose of classification. ITS needs a system that efficiently analyzes the social data in order to meet the needs of traffic control management and mobility users. Therefore, this paper proposes a Latent Dirichlet Allocation (LDA) and word embedding-based sentiment analysis system. The proposed approach retrieves transportation-related data from social media, filters the data to extract valuable information, generates topics using LDA, employs skip-gram technique for word representations, and then uses machine learning approaches to predict the polarity of transportation features.