Enlightening the Bus Oracle - Increasing the Utility of a Bus Information System Based on Natural Language Processing
MetadataVis full innførsel
BusTUC is a Natural Language Processing (NLP)-based system for bus information in Trondheim, Norway. Its service is used in numerous applications, and new applications are still being made. However, not all of these see much use, and users of the system are often only included in the evaluation of the result. Because of this, some BusTUC projects end up having little effect on Trondheim s bus passengers. This thesis details an approach to improving BusTUC that ensures that improvements both reach the users and are perceived as useful by them. A review of the relevant literature was conducted, both establishing the necessary background for BusTUC and exploring recent developments in BusTUC and other NLP systems. Then, a survey was created to investigate which additional functionality there is demand for in BusTUC. With an analysis of earlier surveys biases, the new survey had the soundness of its results increased. Focusing on improvements to the core text-to-text system to ensure that all applications of BusTUC would benefit from them, the thesis identifies several good candidates. The results of this research was analyzed and discussed, leading to the implementation of three of the improvements identified as highly useful. Analyzing the resulting expansion of the system s functionality, the thesis concludes that the approach has merit and does indeed help ensure that the development is beneficial to bus passengers.