Reinforcement learning-based annotation for Deep Web data

A semantic annotation method for web database query result which is under the condition of uncertain schema information is proposed in this paper by adopting the reinforcement learning method. Using domain ontology to annotate the query result, this paper takes the mapping between domain ontology and query result as a process of finding the best strategy. By training the numeric attribute value, we can find the best strategy to match and then annotate the query result. By collecting web databases from different domains, the experiments indicate that the approach proposed can annotate the web database query result properly and improve the efficiency of annotating.