|dc.description.abstract||Background : It can be daunting to begin the search process for a complicated question, particularly if a searcher is new to a topic and not yet familiar with the language used to describe it. As many biomedical searches yield several hundred citations or more, it would be helpful for a user to find some way to summarize her results and see the context into which individual search terms fit. This describes an attempt to meet this need by adapting the Word Trees tool to visualize PubMed search results.
Description : Word Trees were developed by Wattenberg and Viégas, based on their work on IBM’s Many Eyes platform. Much like Keyword In Context, Word Trees present terms in the context of the words that adjoin them. Uniquely, Word Trees also show the relative frequency of these combinations in an intuitive way.
The application described here takes a user’s input and interactively performs a search against PubMed using NCBI’s E-utilities API. It then extracts the titles and abstracts from the first hundred search results and graphs that text using the version of Word Trees that is hosted on Google Charts. This process generates a visualization of the context in which individual terms appear for that set of results. This in turn can provide insight into how those terms are used in the broader biomedical literature.
Conclusion : Simple searches show expected results -- “breast” is commonly followed by “cancer”, and those two words are most often succeeded by “cells”. However, even simple searches can yield insights if one takes advantage of the interactive nature of the Word Tree interface. A search for “morphine” shows “equivalent” as a common succeeding word, and selecting that quickly leads one to the term of art, “MME”. This demonstrates how a searcher using this tool might uncover useful terms for search strategy refinement even before she begins examining her first set of results.||en_US