Facility managers can significantly benefit from operational data, such as maintenance requests, stored in computerized maintenance management systems (CMMSs). This data is a valuable means to assess building performance and gain insights for preventive maintenance actions. However, databases are not always organized in such a way that allow undertaking analytics, therefore resulting in troubles when trying to generate useful information from raw data. This paper presents two methods based on a text-mining approach to extract valuable information from textual maintenance requests. The first method aims to extract the room identifier (ID) numbers where faults mainly occur, while the second one aims to identify the most problematic building elements and systems. The text-mining-based methods were tested by using a data set which contains 12,655 maintenance requests derived from a cluster of 33 buildings managed by the local administration of the Municipality of Trieste (Italy).
