Emotion-based social computing platform for streaming big-data: Architecture and application

Exploration of user generated content in the epoch of Web 2.0 brings unprecedented challenge to the social computing, which has to provide real-time solution in the circumstance of massive data volumes and evolving application scenarios. This paper presents an emotion-based social computing platform namely ESC for streaming big-data. The main aim of ESC is to provide sentiment analysis as the foundation of social computing and enable both real-time computation on streaming big-data and batch computation on off-line big-data with high performance and low risk. Different from conventional data processing technologies, ESC is designed as a scalable and QoS-optimized adaptive platform for developers to only focus on business models instead of being distracted by details of the computing infrastructure. In addition, continuous streaming computing is emphasized in ESC to keep tracking on long term dynamic evolution in social media, which can provide a valuable proxy for in-depth social analytics. The architecture of ESC is implemented by distributed storage, sentiment analysis, data parallelism and routing, real-time streaming computation, batch computation and distributed machine learning. And the evaluation results from real-time and batch computations testify the high performance and scalability of ESC. Moreover, a few applications based on it further demonstrates its usability in enacting on different streaming big-data and variety of social computations.