Abstract: Semantic annotation plays a significant role in Semantic Web rescarcho There arc many annotation methods for unstructured documents today. However,none of them takes notice of the fact that the knowledge locates in documenu sparsely, and few of them make use of the structure of a document effectively, which results in that they cannotannotate document well in case that the quality of the document is poor. In this paper, we proposed a Semantic Annotalion Method based on Sparse Coding(SAMSC) for unstructured data. This method starts from initiation by identifying some semantics described in documents by ontology; secondly, in order to determine the correlation between a document and a semantic topic described in ontology, it resolves the paragraph structure and topics of the document iterativcly;finally, this method annotates the documents in the global range of all documents by minimizing loss function. The experiment results demonstrate the performance of this method annotates unstructured documents well in the Web automatically and effectively. Also, it annotates low quality documents better than other methods.

Key words: Ontology, Semantic annotation, Text paragraph structure, SAMSC