Detection of Real-Time Intentions from Micro-blogs

Micro-blog forums, such as Twitter, constitute a powerful medium today that people use to express their thoughts and intentions on a daily, and in many cases, hourly, basis. Extracting 'Real-Time Intention' (RTI) of a user from such short text updates is a huge opportunity towards web personalization and social networking around dynamic user context. In this paper, we propose novel ensemble approaches for learning and classifying RTI expressions from micro-blogs, based on a wide spectrum of linguistic and statistical features of RTI expressions (viz. high dimensionality, sparseness of data, limited context, grammatical in-correctness, etc.). We demonstrate our approach achieves significant improvement in accuracy, compared to word-level features used in many social media classification tasks. Further, we conduct experiments to study the run-time performance of such classifiers for integration with a variety of applications. Finally, a prototype implementation using an Android-based user device demonstrates how user context (intention) derived from social media sites can be consumed by novel social networking applications.