Drug Information Journal - March 2009 - (Page 118) 118 MEDICAL INFORMATION Coccia, Gabutti, Gualtieri, Sora The early retrieved results are PubMed brief citations—volume, issue, and page numbers are missing—but a full and direct link to the database is offered. The results are clustered as follows: • Important words or phrases used in the titles and abstracts of the retrieved articles (displayed in order of frequency) • Important words (those recurring frequently in the articles) • MeSH terms • Terms that have recently appeared in PubMed • Authors’ names • Affiliations • Journal names (highlighting how many articles a given journal has published on the topic and suggesting other journals in which the researcher might be able to publish) • Years of publication through a combination of statistical and linguistic analysis of the contents. Queries are processed impressively fast, and results are more relevant if the queries are constructed using PubMed MeSH terms instead of natural language. The NLM chose Vivisimo’s clustering technology to enhance search functionality of PubMed, and starting in October 2007 the search engine software was adopted by two NLM websites, MedlinePlus and NLM Library. HUBMED HubMed (28) was created by Alf Eaton, who has joined the prestigious journal Nature. The aim of this project is clearly stated on the HubMed homepage: the tool is intended to provide “an alternative interface to the PubMed medical literature database.” The homepage is essential: it shows only one text box to enter search terms and offers a warning that, in our view, is a weak point: Boolean operators are supported, but it is possible to enter only single words without field tags ([MeSH], [Ti], [Ab], [Rn]). However, the versatility of the instrument becomes apparent as soon as the results are retrieved. Results can be displayed through temporal graphs, two-dimensional categories, or meaningful clusters (related articles, expand search terms). They can be exported in different formats and stored with special tags (personalized annotated lists). Noteworthy is the Rank Relations function. Once the user has selected the most interesting citations, HubMed compares all these citations with its own related articles list and ranks results on the basis of their relevance. Users can employ this list repeating the Rank Relations process until a sufficiently relevant list has been compiled. Overall, this is a valid instrument although still experimental. Simplicity is one of its best features. PUBFOCUS PubFocus (29), developed by researchers Plikus, Zhang, and Chuong at the University of Califor- A “cluster by topic” command is also present, which sorts results into one or more groups or clusters, depending on the MeSH terms assigned. The searches can be further filtered by word and by topic, according to semantic categories (eg, activities and behaviors; anatomy; chemicals and drugs). Anne O’Tate’s response times are not particularly short. This instrument can manage searches up to 5,000 results, but it does not offer an online help service, nor is it possible to find an online tutorial. VIVISIMO Vivisimo (27) is produced by a private company founded in 2000 by a group of research computer scientists from the Carnegie Mellon University in Pittsburgh, Pennsylvania. The company’s aim is to provide users with diversified tools that can face information overload on the web through an information aggregator, thus saving time and generating more accurate results. Among its initiatives, the company has applied the Velocity Clustering Engine to PubMed to create a service it calls ClusterMed. This tool allows subscribed users—a free version can also be downloaded—to cluster up to 500 results per PubMed query. The clustering is achieved
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.