ClusterMed organizes the long list of results returned by PubMed into hierarchical folders with meaningful categories, allowing researchers to hone in on the most relevant results quickly. ClusterMed does this on-the-fly without requiring any pre-processing, using terms taken from the brief descriptions in the search results. With the folders, users can discover themes, view related articles, and drill down from the topic hierarchy or drill up (show-in-clusters) from a specific article. Vivísimo calls ClusterMed a breakthrough improvement in search productivity and efficiency. Users will call it an answer to information overload.

PubMed is the search service of the National Library of Medicine that searches MEDLINE and other databases at the National Institutes of Health (NIH). ClusterMed sends users' queries unchanged to PubMed, and the PubMed search syntax is preserved. Clustering is performed on the text in the article titles, abstracts, and MeSH terms (when available) within PubMed's "pseudo-XML" output.

• Title, Abstract, MeSH terms
• Title and Abstract
• MeSH terms only
• Author
• Affiliation
• Date published

ClusterMed incorporates proprietary biomedical knowledgebases and algorithms for the sophisticated text processing of the PubMed records. The linguistic processing makes use of "semantic, syntactic, morphological, and orthographical knowledge specific to science and biomedicine." Raul Valdes-Perez, president and co-founder of Vivísimo, said the company had spent considerable time building the knowledgebases, such as including equivalencies for terms. Vivísimo has built upon its experiences of working closely with several of the largest pharmaceutical firms, and with biomedical journals and societies.

Valdes-Perez said the company plans to do similar projects with other government sites that have good XML output. In fact, it has already done a demo of clinicaltrials.gov, an NIH site, but hasn't yet decided what to do with it. If you are interested, take a look at: http://www.vivisimo.com/demos/clinicaltrials. Clustering can be done on Topics, Condition, or Status (recruiting, suspended, completed, etc.).

Actually, there are many other interesting demos available at Vivísimo's site (http://www.vivisimo.com/demo/Overview.html), such as the demo that clusters all Microsoft patents (titles and abstracts) issued since 1996 by the U.S. Patent and Trademark office. Vivísimo openly demonstrates its products on various public information sources, such as the Web (check out the Yahoo! News clustering), corporate (Boeing, Merck, HP, etc.), government search engines (FirstGov), scholarly publications, and even eBay.

The company is definitely looking to fill some needs in other data-hungry environments, though Valdes-Perez could not provide further details at this time.

ClusterMed is available for Linux, Windows, Solaris, and other UNIX systems. The company claims that one mid-priced server should suffice for even large organizations. According to Valdes-Perez, yearly subscription pricing would range from about $10,000 for a small company to a high of about $99,000 for a large company. He said the company was also exploring the possibility of an ASP model for the product.

Headquartered in Pittsburgh, Vivísimo was founded by research computer scientists from Carnegie Mellon University. Vivísimo's current customers include MicroPatent, Stanford University's HighWire Press, Institute of Physics Publishing, NASA/Battelle, Cisco Systems, and many others.

Vivísimo's Clustering Engine is currently a finalist (along with IBM and Microsoft!) for the eWEEK Excellence Award in the category of portals and knowledge management. Vivísimo just could be the "little engine that could."