Ranking with tagging as quality indicators

As more and more Web documents are getting tagged, using tags to enhance the quality of retrieval has become an emerging challenge in the retrieval. In a clear contrast to traditional retrieval techniques solely focusing on content-based relevance to query keyword, we aim at enhancing the retrieval with additional contextual information obtained from tags, e.g., popularity or reliability of the document. Toward the goal, this paper views a tag as a "quality indicator" and classifies tags with various characteristics adopting the taxonomy proposed in data quality modeling literatures. We then develop a unified framework for supporting arbitrary quality indicators in the taxonomy and evaluate its effectiveness and efficiency over both real-life and synthetic data.