Railway Complaint Tweets Identification

Several government ministries have been using social media platforms, mostly Twitter, for information sharing and complaint gathering. This opens the door for collecting information, knowledge, sentiments and needs of the citizens and applying analyzed citizen sourcing ideas to provide better public services. In this work, one such usage of Twitter by the Ministry of Railway, India is analyzed. A large number of tweets are received on Railway Ministry official twitter handle in an hour having suggestions, complains, grievances and opinions about railway services and issues. It is important to identify the complaint tweets for the Railway Ministry to take further actions. Due to the unstructured nature of tweets and high velocity, it is laborious to identify the complaint tweets. The complaint tweet identification problem is modelled as social network graph using tweet reply feature. After exploratory analysis of the obtained graph, several linguistic and non-textual tweet features are identified and two SVM classifiers with linear kernel are used to classify complaint and suggestion tweets. The classifier achieves the classification accuracy of 92.09 and 70.62% for complaint and suggestion tweets, respectively. Complained tweets are further analyzed to identify the types of complaints and frequently co-occurring terms using topic modelling.