Keep up with all of the essential KM news with a FREE subscription to KMWorld magazine. Find out more and subscribe today!

Search Technology: Greater power, but increasingly simple to use

Article Featured Image

The new generation of search technology is characterized by the use of AI and machine learning, natural language processing capabilities, and much greater analytical power than has been previously seen. In general, though, these products are being combined with improved usability, so that despite the increased power of this more sophisticated technology, the software has not become more complex to use; on the contrary, it is more responsive and easier to use than previous products. Particularly in the case of cloud solutions, the support provided by vendors is removing considerable burdens from the organizations that deploy the solutions.

Identify requirements

UW Health is an integrated health system with six hospitals and 80 outpatient sites serving over 600,000 patients per year in the upper Midwest. The system has 1,500 physicians and 16,500 staff employees. Affiliated with the UW School of Medicine and Public Health, UW Health provides patient care, research, education, and community service missions. Its intranet, referred to as U-Connect, is the central repository for clinical information, documentation, and information about the organization that is needed by employees.

“UW Health has a large volume of content, a great deal of which is unique to the organization,” said Noah Locke, manager of web technology. “Being able to find critical content is essential, and each group has very different needs; for example, the nurse in a burn unit is looking for different material than a pharmacist would be seeking, so being able to do precise searches is very important.”

Over the years, UW Health has used a variety of approaches to searching, including two versions of search appliances from Google, an open source product, and a proprietary search engine. Although each one achieved some degree of success, each also had limitations or problems, ranging from the discontinuation of the appliances in two cases, to maintenance demands, or not meeting the expected performance levels.

Having opted not to renew with its latest search software company, UW Health was under pressure to find a new solution quickly. “We were in a situation where we had only a 6-week window to check out options, see demos, and make a choice before a 30-day implementation period,” Locke noted. After seeing a demo of the administration console of the Coveo cloud search platform, he was certain that it was the best match for their requirements. “It had the specific functionality we needed and was within our price range,” he added.

Key capabilities

The implementation, which took place unannounced, went very smoothly. “We wanted to see if people noticed the difference,” commented Locke, “but we did not get any complaints. In fact, the results were as good on the first day as they had been when we were using the search appliance, and in that case, we had been applying machine learning for quite some time.”

One type of content that was particularly challenging was a set of “Health Facts for You,” referred to as HFFYs. “These are one-page patient education documents, and we have several thousand,” Locke explained. “Because they are PDF documents, they can’t be indexed the way a Word document would be. Also, they are updated frequently. Our previous search tools needed very specific rules for crawling this set of documents. With Coveo, all these problems vanished, and the updating took place quickly and accurately.”

The analytics capabilities in Coveo were also a pleasant surprise. “We did not have this level of insight with our previous products, which were producing rather complicated reports and spreadsheets,” observed Locke. “We began getting actionable information from Coveo right away.” The statistics indicate how long people are searching without finding what they need, or what the click-through rate is, which reveals whether the target document was found. “We can also look for content gaps, which with Coveo happen only rarely because the right content is usually found. But if not, we can tweak the search technique or can provide additional content.”

KMWorld Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues