Director's Blog

Google has become the gold standard for search - with good reason. It works. Really well. Google makes all of us smarter by surfacing relevant information, quickly. Were it not for Google search, I'd probably still be looking for information on that cool new cgi-bin technology and checking my Eudora mail account.

I try to avoid Endeca versus Google comparisons because - for OHSU's needs - they really aren't comparable products. But on the surface, the core function looks the same: submit term(s), receive links. And while we strive for that same effortless return of perfect results, Endeca is retrieving information from a very different pool of data than Google is working with, which produces very different outcomes.

Rather than write an epic novel about these outcomes where I pretend to have all the answers about what happens and why, I'm going to occasionally respond to common questions that I do have a reasonable response for.

Why does Google do so much better at surfacing home pages than Endeca?

Search engines commonly base their relevancy ranking strategies on the number of links that point to any given page. Internal search engines don't have this external data available to the ranking modules, nor would it be a particularly reliable source of information to define a home page if it were based solely on internal links - which means we really have no organic way of defining which pages are "home pages."

I'd love to find a way to mimic this behavior with Endeca. I'm in the midst of testing a custom ranking strategy to float web site home pages higher up in the results. I'm starting with CommonSpot subsites because we have fast, easy access to tag home pages with the right meta data. I'd be happy to throw any other site into the early test phase though - send me an email if you'd like to participate.

Feel free to send me any of your Endeca v. Google questions.
Renee

OHSU's websites use several different search technologies - we have everything from enterprise software (Endeca, Verity) to free tools (Google search, PHP scripts) to custom-built solutions.

In many cases, there's really no way of knowing which tool is powering any individual interface. To set the record straight, what follows is an explanation of some of our most frequently used search applications.

Endeca - Enterprise Search
http://www.endeca.com/enterprise-search/
Endeca powers search for many of our public-facing sites. These search results display on pages with "www.ohsu.edu/searchresults" in the URL. If the page has a different URL, it's not Endeca.

HtDig - Ozone Search
http://www.htdig.org/
The Ozone search tool is a free application that desperately needs to be replaced. As far as I know, the configuration hasn't been touched since it was installed more than four years ago. Unfortunately, the Ozone didn't make the cut for projects being supported after budget reductions.

Custom - OHSU Health Website
The health topics, find a doctor and find a clinic search features on the health website are custom-built. There are no third-party applications involved.

As always, I welcome your feedback.

I've heard a rumor or two (or three, or four, or...) that our internal search engine is not quite functioning like we need it to. I'm looking for detailed feedback on Endeca searches that aren't working.

The best format for this feedback is:

• search term (you can even copy the URL after you run the search; your term will be in there
• desired URLs in the top 10 search results
• other notes about the search

For example:

• search term: "school of medicine"
  http://www.ohsu.edu/searchresults/controller.jsp?Ntt=school+of+medicine&Ntk=All&rpp=10&search=Search+Again
• desired URLs: www.ohsu.edu/som, http://www.ohsu.edu/ohsuedu/academic/som/dean/deans-message.cfm

Send this info to webadmin@ohsu.edu or open a ticket via https://www.ohsu.edu/webstrategies/support.cfm.

Look forward to hearing from you!

The search engine upgrade is complete!

This upgrade brought our installation of Endeca up to its most current version. We also launched an updated search interface based on the new Unified Site HTML and CSS templates.

There were significant improvements made to the site crawler in this version of the software. Endeca changed the name from "advanced crawler" to "content acquisition system" - and it looks like there's more to the upgraded technology than just a fancy new name. Our internal testing shows an improvement in accuracy of search results.

There are a number of ways your website can be featured in the search results - read more about content spotlighting at: http://www.ohsu.edu/webstrategies/webtools/search.cfm

Happy new year!

We're in the midst of improving the enterprise search interface (www.ohsu.edu/search) - just a few minor things to make it more usable:

• Ability to apply/retain mission color formatting depending on where the site is launched from
• Adding icons to indicate document types, like PDF, Word, etc instead of the words following the title
• Dynamically displaying and hiding the right column so it never shows up empty
• And my favorite, "type ahead" search, so Endeca will make suggestions for searches as you type

Brenna Harris is my hero for identifying the cause of an IE bug that was holding up development. Who knew that comments before the doctype declaration will convert the page to quirks mode? Not me. I do now. Can I have those four frustrating troubleshooting hours back please? If you're curious, the complete explanation can be found here: http://reference.sitepoint.com/css/doctypesniffing

I'm hoping to launch the new interface mid-December. Feel free to email me with any requests or suggestions.

A couple of months ago, I told you that our new search engine was only in its first phase. Phase II will become available later this month. This will address several of the issues we've been hearing about.

The first enhancement is advanced content spotlighting and comes in four parts. First, you can now request a keyword, phrase or set of keywords and specify a web resource for them to be linked to. When someone queries on those terms, a link to your page and a brief description will be displayed at the top of the results page above the rest of the query results.

In addition to featuring a link at the top of the result set, a set of keywords can trigger the display of a promotional image in the left hand column or featured text in the right hand column. You define and provide the information that will be displayed whether it be an image that links through to a web page or the title, description and text of a snippet you wish to promote.

Finally, content spotlighting will now include "related news" items in the left hand column. For example, if you search on "MRI" both clinical trials such as "MRI Evaluation of Esophageal Cancers (MRI-EC)" and relevant news items like "OHSU Receives Two-Of-A-kind MRI Magnet" are displayed.

The second big step forward for search at OHSU is the availability of subsite search.. Endeca can now be used to search any collection of directories or websites you specify. The search box will look and function exactly the same as the main search box, however the search results will only include pages and resources found within the subsite you define. Web Strategies will help you define your subsite, register it within Endeca and get a new subsite search box on your web pages. To get started, fill out the subsite search request form which will be available shortly on the web strategies website.

As units across the university take advantage of this new capability conflicts and competition for certain keywords and phrases are bound to arise. Web Strategies will work with units and the marketing and communications group to ensure that any conflicting requests are resolved in a manner that meets the needs of both the unit and university as a whole.