FRAMINGHAM (04/05/2004) - Vivisimo Inc. has launched ClusterMed, a search tool designed to improve the efficiency of PubMed database article searches.
Specifically, while a native PubMed search returns a chronological list of all articles matching a search criterion, ClusterMed automatically groups the articles into hierarchical folders based on subject categories.
To sort articles into relevant folders, "ClusterMed takes advantage of the XML output and structure within PubMed documents," says Don Taylor, Vivisimo's senior director of life sciences solutions.
The basic idea behind ClusterMed is to give researchers a way to quickly identify themes within a list of articles. For instance, a search for "protein folding" using the limited function ClusterMed demonstration on Vivisimo's Web site produced about 40 clusterings with categories such as protein-folding kinetics, NMR studies, Escherichia coli and cross-linking.
For some categories, a second-tier clustering is produced. For instance, the Escherichia coli folder sorted articles into nine sub-topic areas, including two-state folding, binding activity, and crystal structure.
Nuts and bolts
When conducting a search, a researcher enters the words or phrases of interest. The search produces a split screen, with the hierarchical folders in a viewing pane on the lefthand side of the screen and the list of articles in a pane on the righthand side of the screen.
The articles pane displays basic information about each article, including title, abstract, a list of authors, the affiliation of the authors, the full article reference, and the article's URL.
Naturally, many articles may end up in multiple category folders. A simple click on a "show in clusters" link (which is at the end of every article listing) will highlight all the folders (in the lefthand cluster folder pane) that the article has been sorted into.
ClusterMed allows articles to be sorted by date, MeSH (medical subject heading), author, or affiliation. Vivisimo says sorting articles by author or affiliation can be used to help identify experts or the leading institutions conducting work in a particular area of research.
A demonstration version of ClusterMed can be found on Vivisimo's Web site. Unregistered users of the demo can get up to 100 articles returned per search; registered users get a 30-day trial that supports up to 500 results per query.
Vivisimo offers site licensing for the full-blown version of ClusterMed, which is available now.
The ClusterMed tool is based on Vivisimo's search and clustering technologies, which are used in the company's more general content-, knowledge-, and information-management products.