The electronic product code (EPC) network is an important part of the internet of things. The physical mark-up language (PML) is to represent and describe data related to objects in EPC network. The PML documents of each component to exchange data in EPC network system are XML documents based on PML Core schema. For managing theses huge amount of PML documents of tags captured by radio frequency identification (RFID) readers, it is inevitable to develop the high-performance technology, such as filtering and integrating these tag data. So in this paper, we propose an approach for measuring the similarity of PML documents based on Bayesian network of several sensors. With respect to the features of PML, while measuring the similarity, we firstly reduce the redundancy data except information of EPC. On the basis of this, the Bayesian network model derived from the structure of the PML documents being compared is constructed.