A new approach for train calendar description generation

This paper describes a new method of generating text starting from a calendar, automatically and in a representation clear to customers. We focus in particular on railway applications. Railway undertakings are trying to improve their communication with commuters, employees and travelers, especially for the frequent occurrences of path modifications implying scattered calendars and not intelligible descriptions appearing in various outputs such as websites, mobile applications, timetable boards, train transport diagrams and books. We propose two alternative approaches for this challenging task. The first one combines a customized set covering problem formulation with a parallel vector generation algorithm. The second one integrates the vector generation problem in the set covering problem into a more complex mathematical model. Our aim is to verify that with a mathematical programming approach it is possible to improve the quality of outcomes in terms of intelligibility. We used a commercial mip solver (CPLEX 12.4) to solve the two models, while we designed a specific parallel algorithm for the vector generation problem in the first approach. The new solutions were tested on several real timetables and compared with the tools used so far.