Celia Boyera, O. Baujardb, V. Baujarda, R. D. Appela, b, c
a Health On the Net (HON)Anyone using Internet needs, at some point, a search engine to browse or search a subject area or to locate a specific piece of information. However, anyone using a search engine has experienced that the results obtained do not always meet the request. It is often necessary to review a long list of sites, and not always evident to pick those that most closely match one's quest. A solution to these inconveniencies has been proposed by the Health On the Net Foundation (HON) a few years ago with MedHunt [1, 2], the medical and health dedicated search engine and the associated robot, M.A.R.V.I.N (Multi-Agent Retrieval Vagabond on Information Networks) developed in collaboration with the Swiss Institute of Bioinformatics [3]. MedHunt helps patients and the medical/health community find the required information. The utility of such a tool is expressed by its growing use with 90,000 accesses per month and 1,200 links to its URL [4]. Over the last years developments in information retrieval were more and more centered on the new concept demonstrated by M.A.R.V.I.N which is based on a specialised domain specific dictionary, a parallel multi-agent structure and filtering processes and techniques [5, 6]. The technical concepts having proven their robustness, new developments focus on the ergonomic interface and the practical use of a search. This article presents the new possibilities of MedHunt that support and facilitate the search for relevant medical information.
Few users know how to obtain the information they need, and most Internet searches are misspellt or misstyped. Practical utilities aiming at providing user-friendlier interfaces and searches were designed in 1998 and have been integrated to MedHunt. The new developments, detailed below, associate:
Spell checking is performed on the medical and health terms included in MedHunt's dictionary. If a query returns no results, then MedHunt proposes dictionary terms that have a similar spelling (at most four spelling differences). For example, by entering "osteoroporosis", the spell checker displays osteoporosis, osteoporosi, osteoartrosis, osteoarthrosis, osteonecrosis, etc. The user can then redo the search by clicking on any of these terms.
This utility is available when the returned list of sites includes highly scored documents. The Medical Topics that are displayed are the twenty most occurrent medical/health terms automatically retrieved from the highly scored documents. For "osteoporosis", the Medical Topics proposed are: osteoporosis, medical, health, disease, menopause, postmenopausal, calcitonin, hormone, vertebral, arthritis, diabetes, etc.
The Medical Subject Headings consist of a classified and controlled medical vocabulary developed at the National Library of Medicine [7]. MeSH organizes its 18,000 concepts into a hierarchical structure including anatomy, organisms, diseases, chemicals and drugs, psychiatry and psychology, etc. and provides the description of these terms [8]. The MeSH available on-line has been entirely downloaded and is used through the MedHunt service, MeSH description, thus enabling medical/health queries to be performed using MeSH terms [9]. The subset of the hierarchical structure that includes the request is thus displayed. The user can then browse the structure in order to find appropriate concepts and then perform a new search with MedHunt [1], Medline [10] or Yahoo! [11] to find Web sites, article references and abstracts, or related daily news respectively. Searches for a broad concept within the MeSH description utility can therefore include web sites, articles, or information indexed to narrower concepts. See the example on "osteoporosis" in [12].
This service offers a simultaneous translation of medical queries into eight of the European Union languages: Danish, Dutch, English, French, German, Italian, Portuguese and Spanish. For each translation a new search with MedHunt can be automatically launched.
This feature sorts the results by domain name, continent and country (states for the U.S.A.), and specifies, for each of the selections, the number of sites found by scoring group. This service allows one to choose the result's visualization according to the best score and the geographical location. This is especially useful when a user prefers to visit a site nearest to his/her physical location.
With its new functionalities, MedHunt offers a much user-friendlier approach to medical information retrieval, both for medical professionals and for individuals. It gives access to a wider range of information, and limits spelling mistakes in queries. Also, through its integrated categorization and hierarchical paradigm, it allows for more efficient targeting and deeper refinement of the search procedure. There is thus a higher likelihood that the obtained information satisfies the request.
[1] http://www.hon.ch/HONsearch/Patients/medhunt.html
[2] C. Boyer, O. Baujard, V. Baujard, S. Aurel, M. Selby and R.D. Appel, "HON Automated Database of Health and Medical Information," International Journal of Medical Informatics, Vol. 47, Iss. 1-2, pp. 27-29, Elsevier Science Ireland Ltd., 1997
[4] O. Baujard, V. Baujard, S. Aurel, C. Boyer and R. D. Appel, "Trends in Medical Information Retrieval on Internet," Computers in Biology and Medicine, 1998. In press.
[5] O. Baujard, V. Baujard, S.Aurel, C. Boyer and R.D. Appel, "MARVIN: A Multi-Agent Softbot to Retrieve Multilingual Medical Information on Internet," MedNet'97, Nov. 1997, Brighton, United-kingdom.
[6] O. Baujard, V. Baujard, S.Aurel, C. Boyer and R.D. Appel, "A Multi-Agent Softbot to Retrieve Medical Information on Internet," Medinfo'98, Aug. 1998, Seoul, Korea.
[7] http://www.nlm.nih.gov/mesh/MBrowser.html
[8] http://www.nlm.nih.gov/mesh/MBrowser.html
[10] http://www.ncbi.nlm.nih.gov/pubmed/
[11] http://search.news.yahoo.com/
[12] http://services.hon.ch/cgi-bin/HONselect?browse+C05.116.198.579