A practical problem that has been more and more acknowledged is how to improve the quality of metadata that is created automatically. The rapid growth of data resource has increased the importance of automatic metadata creation. Many methods for automatic creation about metadata of the documents have been investigated and they are similar in principle but different in details. We propose a complementary approach, based on collaborative annotation information to reflect the opinions about the content in a document from different readers’ view, that achieves automatically-generated metadata enhancement. In our method, we analyse the influence of different features of annotators and give an evaluation of their role in enhance the quality of metadata. It is also demonstrated by our experiments that the result of generated metadata is varied with the annotators’ features, and the result is satisfactory, if the influence of annotators exceeds certain threshold. A practical problem that has been more and more acknowledged is how to improve the quality of metadata that is created automatically. The rapid growth of data resource has increased the importance of automatic metadata creation. Many methods for automatic creation about metadata of the documents have been investigated and they are similar in principle but different in details. We propose a complementary approach, based on collaborative annotation information to reflect the opinions about the content in a document from different readers’ view, that achieves automatically-generated metadata enhancement. In our method , we analyze the influence of different features of annotators and give an evaluation of their role in enhance the quality of metadata. It is also demonstrated by our experiments that the result of generated metadata is varied with the annotators’ features, and the result is satisfactory , if the influence of annotators exceed certain threshold.