Real-time government policy monitoring framework using expert guided topic modeling

The user generated data on social online platforms is a very resourceful medium for building applications. It provide an easy, cheap and efficient way to address the opinions expressed by millions of users. There are various commercial applications developed that analyze the products based on user reviews. However, no considerable research work is carried out at building applications to consume this data for good governance. Topic models are successfully used for various text analysis tasks and are employed here. In order to improve accuracy of the model as government policies is a specialized domain, the model is extended with expert guidance by following a semi-supervised approach. POS tagging is used to identify user opinions which is evaluated through WordNet dictionary. In order to perform real-time analysis, the model is tuned to streaming tweets data which is analyzed to produce real-time analysis about the top government policies. The model is trained on print media articles and is tested on twitter data that accurately recognized policies and represented their sentiment scores for different time spans.