Web annotation has become an essential technique to express people's thoughts by attaching annotations to web content. Annotators can exchange their ideas by conducting universal collaborations. The idea of submitting annotations depends on attaching vocal or textual notes with the contents of websites. However, annotating dynamic data is a problem that encourages researchers to work on proper solutions. Losing annotations because of the change in website annotated contents will definitely lead to losing the intended collaboration between annotators. This work is related to annotating dynamic websites by computing the similarity between erased (or relocated) annotated text and the remained text in dynamic websites by exploiting NLP algorithms. The attached annotation with dynamic content will be attached to the most related text on the website. By this, annotators will not lose their annotations and replies and hence their collaboration will remain. The experimental tests conducted in this work reflect promising results.