Maintenance management forms a core component of building facility management, which is particularly important for specialized and complex healthcare facilities. However, existing maintenance management tends to rely on manual processing, which is time-consuming and human error-prone. Recorded maintenance work orders (MWOs) are recognized as good potential data source to support maintenance activities. In this paper, the authors propose a comprehensive framework utilizing natural language processing (NLP) techniques to automate the interrogation of textual data of MWOs that assist in developing maintenance material management and preventive maintenance strategies and transforming maintenance worker assignment from a labor-intensive process to a more automated one. A set of historical MWOs from a hospital was used for model development, and our model achieved an accuracy of 0.83 on worker assignment. This study addresses the difficulty of utilizing work order information due to the unstructured nature of textual data and contributes to better maintenance management practices in buildings. •Investigated NLP and machine learning models to automate the processing of hospital maintenance work orders in Chinese.•The model is configured with multiple combinations of splitters and machine learning algorithms.•Splitting text input into single characters was found to perform surprisingly well.•The model shows high accuracy on worker assignment tasks that outperformed the state of the art.•The study provides a replicable approach to the improvement of hospital building maintenance management practices.