A semantic tagging system for biomedical articles

With the popularity of Web 2.0, people start using tags to label web resources, some even go further by applying Semantic Web technologies to tags. While normal tags impose too many limitations thus not powerful enough, semantic tags are too flexible to address issues like efficiency and quality. We present a tagging system that is built upon Semantic Web technologies and a semantic element is only chosen if we have a clear objective. The system can be used in PubMed, a database of over 19 million citations of biomedical articles, as an auxiliary system to improve the article locating ability a lot with acceptable trade off. We describe the system architecture and the optimizations apply to the system, then we present test results of several random generated test cases. The test results show that when running on medium scale tag repository, our system has quick amortized response time and acceptable memory usage. At last we present validation and correction strategies to deal with unreasonable tag relations in our system.