Public bus commuter assistance through the named entity recognition of twitter feeds and intelligent route finding

Abstract

Karachi (Pakistan) has recently been subject to violent incidents targeted primarily at civilians. These incidents are problematic for commuters who use the public bus system and who often fail to reach their work organisations due to consequent bus strikes. This series of events leads to considerable financial losses for the transport industry. This study proposes and implements safe and fast around the road (SAFAR) which is an intelligent transport Android application developed in collaboration with the local transport authority of Karachi. SAFAR provides run-time information to bus commuters regarding recent violent activities farther up from the current location of the commuters on their route. SAFAR employs live Twitter feeds to classify the manner, location, and casualty information of the violence. The authors investigate SAFAR's performance offline with three named entity recognition (NER) approaches, namely, supervised, dictionary-based, and integrated (hybrid), and show that the integrated approach has the best performance with a precision of 85%. Furthermore, SAFAR recommends alternate routes to commuters if violence is detected farther up through the A-star (A*) algorithm. An online evaluation of SAFAR with 50 real users gave a precision of ∼85% to identify violence locations. Finally, a subjective evaluation showed that SAFAR's performance is satisfactory.