We propose a new tag recommendation method for the public digit cultural resources we collected in related program. We use LDA and Word2Vec to preprocess the resource and tag, map them to different vector space respectively. After that, we calculate the relevance of each resource and all tags by using Deep Structured Semantic Model. Lastly, range the result of calculation and treat the top N tags as the extended tags of a resource. Result show the good performance of our tag recommendation method.