IDEAS

PubMed Search Assistant is an innovative interface for online searching. The interface makes it possible to construct and modify queries visually. The interface also makes it easy to construct controlled vocabulary-based queries in order to achieve high precision and high recall. The innovative ideas of the interface include:

• To apply the "drag-and-drop" technique to help the user modify search queries.
• To automate query construction process using the concept boxes.
• To extract descriptors (MeSH terms in this case) from the retrieved documents and list them in the order of their occurrence frequencies.
• To display the number of hits automatically throughout the query modification process.

USE OF THE INTERFACE

To user the interface for searching, you first type a query in the query field, then either hit the return key or press the lookup button (the query could be a single word or phrase, or any combination of words with Boolean AND and OR).

Instead of directly showing documents retrieved by the query, the interface will first lookup all the MeSH terms in the retrieved documents, and display the terms either in the order of their occurrence frequencies, or alphabetically.  Buttons are porvided to switch the list from one order to the other.

Now the interaction starts. Click on the list to highligh a term, and then drag the term to any of the three Concept Boxes (or use the arrow keys to move the highlighed terms to the conceptboxes).  The concept boxes are modeled after a typical online search planning sheet, that is, any terms within a box will be ORed together, and the three concept boxes will be ANDed together to form a query. The number of hits of this query will be displayed in the "Documents Found" field. The number of hits will be changed each time a term is added or deleted from the concept box. (To delete a term in the concept box, double-click on it).

Watch the number of hits carefully. If the hit is too small, you may want to OR some terms to expand the query. If the hit is too large you may want to AND one more concept to narrow down the query. If the hit is reasonable, you simple click on the "Show Documents" button to display the retrieved documents.

Finally, an Extra field is provided for you to type in any additional words to add to the query. When needed, you may also use this extra field as a forth concept box by dragging a MeSH term to this field. Also remember that you can start a new query any time in the query field, without deleting terms in the concept boxes. In this way, you can explore different concepts and find more MeSH terms related to your interests.
 

THE TECHNOLOGY

The interface is a Java applet. It communicates with a servlet which sends queries to PubMed. The servlet also provides various functions to extract MeSH terms and process query results. The communicaiton was originally implemented in a RMI server. Because of different implementation of Java technology by Microsoft and Netscape, using RMI creating difficulty to use the program with Microsoft's Internet Explorer. The current implementation should work with both browsers.
 

THE FUTURE

This interface is still evolving. it should be even easier to use when most of the browsers are standardized on Java 2 technology.  Currently, we are planning to conduct some usability studies on the interface and add more interaction functions to the interface.

The interface can be modified to work with other databases with controlled vocabularies. The main customization involved is the term extraction process and search engine CGI specification and/or query specification. If you are running a database and you would like to try to customize this interface to your database, please contract Xia Lin at xlin@drexel.edu.

Please send any comments and suggestions to xlin@drexel.edu. Thank you.
To try the interface, Click here