Design and Evaluation of an Interactive Topic Detection and Tracking Interface

Interactive Topic Detection and Tracking (iTDT) is a branch of TDT focussing on aspects of the user interface and user interaction. The importance of user interaction in the real world is the reason why iTDT is receiving attention. This particular research project has been motivated by the fact that there has been very little exploration of the usability of the features introduced in the iTDT interfaces. This research investigates the successful components and features of an iTDT interface by implementing them into a single interface called Interactive Event Tracking System (iEvent). The usability of the features introduced in iEvent, including the usefulness of Named Entity Recognition (NER), were then thoroughly evaluated. The key features of iEvent Cluster View, Document View and Term View have enabled journalists to perform various TDT tasks. The user tasks in this study are designed to support the journalists to perform TDT tasks in a way that it is in line with their task. Therefore, this is a ground-breaking study which investigates journalists as iTDT interface users. Key findings revealed that iEvent enables journalists to perform well in TDT tasks. The research also identified potential guidelines for future designs of iTDT interfaces for TDT tasks. It was also revealed that, in general, Cluster View is useful and interesting; Document View is effective and interesting; and Term View is helpful. In addition, NER facilitates the Tracking task while keywords are valuable in the Detection task. NER emerges as more valuable in the Tracking task because it can effectively enhance user interactions with the system and brings efficiency to the Detection task. Overall, these findings supported the hypothesis that the use of NER improves iTDT as it improves standard TDT.