To explore the valuable thematic information and sentiment distribution in patients′ medical service complaint texts based on topic modeling and sentiment analysis, and investigate the main driving factors affecting patients′ service experience and satisfaction.

Topic mining was carried out on the offline patient complaint text set of a tertiary hospital in South China from 2013 to 2017. The seed word set extracted from 1 000 sampled texts was used to guide semi-supervised Latent Dirichlet Allocation training of texts. Relevant subject categories were extracted and subject characteristics were graded emotionally.

in the end, 30 subject categories were extracted from the 8 000 complaint texts, and the sentiment score of the subject characteristics was consistent with the sentiment tendency of the actual data set. However, the satisfaction was relatively low in " toilet" , " ward" , " hygiene" and other subjects, and the main complaint subjects included " attitude" , " examination" , " ward" among others.

Based on the theme distribution, combined with the results of emotional analysis and the specific clinical environment, strengthening management in the medical service sector with a large negative emotional score can guide the hospital management practice and service improvement process, and help to improve the patients′ perception experience and emotional experience.

Medical service experience; Satisfaction; Negative text; Topic modeling