Under the trend of deep integration of culture and tourism industry, tourism products have become a new type of carrier to reproduce the value of non-heritage culture. On the digitization of the inheritance and innovation of non-heritage Manchu embroidery, this paper proposes a total of two effective paths to collect and publish the digitized information of Manchu embroidery culture and establish a database. And the graph structure method is chosen as the mathematical modeling method for the information of Manchu embroidery culture. Unsupervised pre-training is carried out based on the topological features of the graph dataset to build a pre-training model that can capture the features of contextual information. On the basis of this model, fine-tuning based on the graph pre-training model is proposed to build a model framework for feature migration. At the same time, a platform framework containing the functions of the three modules of resource collection, resource retrieval, and user participation is designed, which is combined with a database to form a database platform for the non-heritage Manchu embroidery cultural heritage. The designed database platform method has an average performance of 0.904 in terms of accuracy rate in information extraction performance, which can provide powerful data and information support for the construction of non-heritage full-embroidery knowledge map as well as the design of tourism products.