refmed

Related Articles

• H Yu, Selective Sampling Techniques for Feedback-based Data Retrieval, Data Mining and Knowledge Discovery, Springer, Volume 22, Issue 1 (SCI),  2011.

• H Yu, T Kim, J Oh, I Ko, S Kim, WS Han, Enabling Multi-Level Relevance Feedback on PubMed by Integrating Rank Learning into DBMS  BMC Bioinformatics, Springer, 11(Suppl 2):S6 2010.04. (SCI)

• H Yu, T Kim, J Oh, I Ko, S Kim, RefMed: Relevance Feedback Retrieval System for PubMed  ACM Conf. on Information and Knowledge Management, 2009. (CIKM'09) 

Demo site

• http://dm.postech.ac.kr/refmed/

Get started

RefMed provides the relevance feedback search function for PubMed: When a user enters a keyword in the search box, the PubMed search results will be returned. The user then specifies on a sample of results how much they are relevant to what she intends to find, for example, by specifying whether each article is high relevant, somewhat relevant, or not relevant. Once the user clicks "Push Feedback" button, the system learns a relevance function from the feedback and returns the top articles ranked highly according to the relevance function. The user can repeat the process until she gets satisfying results.

To illustrate, enter a keyword "mrsa" in the search box and click Go button.

The figure below shows the results. (1) There are three-level radio buttons in the right side of the results on which the user specifies the relevance degree. The user may specify the relevance for any articles and don't need to evaluate the entire results. (Typically, more relevance feedback would increase the accuracy of the relevance function.) (2) The user clicks "Push Feedback" button to proceed.

Below shows the results returned by the system, which ranked highly according to the relevance function that is learned from the user's feedback. The user may keep providing feedback to improve the accuracy of the relevance function.