Abstract: 【 Purpose/significance】This paper accurately measures the user emotional experience of library intelligent service through multimodal emotional analysis, and propose feasible countermeasures and suggestions for library intelligent transformation.【Method/ process】The attention mechanism and LSTM algorithm are comprehensively applied to build a multimodal emotion analysis model in⁃ tegrating text, voice and expression features, and the validity of the model is verified through MOSI open data set. On this basis, the ro⁃ bot intelligent question answering service of Wuhan University Library is taken as the empirical object to measure the user′s emotional experience【. Result/conclusion】The application of multimodal emotional analysis model can comprehensively and accurately measure users′ emotional experience, and reveal the key factors affecting users′ emotional experience effectively, thus providing an important basis for emotional design and experience optimization of library intelligent services【. Innovation/limitation】This paper proposes a mea⁃ surement method of emotional experience of library intelligent service users based on multimodal emotional analysis, which provides a new idea for the library to accurately evaluate the user emotional experience under the digital intelligent environment. Future research can integrate more modal features to further improve the performance of the emotion analysis model, and conduct empirical research in more intelligent service applications.