Complex Search @ Your Web Service

Page 3 of 3


Good or Just Good Enough?
Gilbane points out that no matter how good a tool may be, there are always going to be search repositories that, for whatever reason, lack good metadata. At that point, he says, you become dependent on full-text search to go that last mile. Gilbane adds that there may be some applications where this type of searching won't be good enough to get you what you need. "In healthcare applications, for example, in a medical situation or emergency where you are looking for a particular medication or drug you need quickly, you can't depend on full-text search, you need something more rigid. If you were doing research or writing a paper, it wouldn't be as critical, but if it is a life-threatening situation, you need to be absolutely sure

No matter how good the conversion tool is, according to Gilbane, conversion doesn't always parse the content as well as it should. Thus, search tools won't be able to make it granular enough, regardless of XML conversion.

Certainly, companies can construct XML data stores where none exists, or build an XML-based cache of critical data. In many cases, though, creating data stores from whole cloth will not be as feasible as converting existing ones to XML. However, even without the most perfect conversion, Web Services-enabled searches give enterprise users the ability to take data results and better incorporate them into the workflow cycle and real business applications. This not only provides a methodology for producing a more useful result set, but a mechanism to better incorporate found information into producing meaningful results—and that is something all businesses are searching for.

Sidebar: New England Journal of Medicine Uses Mark Logic Technology to Improve Search

Kent Anderson, executive director of international business and product development at the New England Journal of Medicine, wanted to make better use of the Massachusetts General Hospital (MGH) case records. This repository was used by students and instructors to provide sample cases for teaching purposes.

Users were not able to pinpoint the cases they wanted. What they had in their case repository, according to Anderson, "had been popular for decades, but there was no way to mark up and extract the really salient points, the presenting complaints, what somebody shows up with, and then the final diagnosis. Now we can do that."

Anderson says he had a discussion with Mark Logic and was impressed with what they had to offer. "We had been looking around, but Mark Logic came up very early in our examination and seemed to have the right attitude and technology." When the Journal implemented the Mark Logic solution, they added a tab in their search tool to search cases, and this enabled students and instructors to isolate and retrieve cases in the MGH repository based on different data types such as the diagnosis, the tests given, and other specific information. Users can quickly identify the type of information using an icon system on the search results page.

Anderson says that before they implemented this new system, they would just get a list of cases without any context or understanding of which tests were given or how they arrived at the diagnosis. With the old system, users could enter a complaint or a diagnosis such as heart disease, but just got a list of results in which the term was mentioned. In the new system, he says, users can isolate results around specific issues such as the type of test, final diagnosis, or patient complaints.

He says end users love it because they can pinpoint the cases that match their needs with far greater ease and accuracy than was possible under the old system. "Physicians-in-training have found it remarkable."

Companies Featured in This Article

Alexa Web Information Service
Bluebill Advisors
Ipedo Mark Logic
New England Journal of Medicine
Rollyo Siderean
Yahoo! Developer Network

Page 3 of 3